INTRODUCTION
While crime prevention and reduction remain key concerns for most countries, there are certain types of crime which, despite their relatively low frequency (compared to other types of crime), have a disproportionate impact on social order, fabric and economy, namely collective violence (Gaibulloev and Sandler Reference Gaibulloev and Sandler2019; Waters et al. Reference Waters, Hyder, Rajkotia, Basu and Butchart2005). The collective violence label relates to crimes motivated by the identification with a given collective and against members of another collective because of their membership with said collective. Collectives can be based on ethnic, religious, national, political or group-based identities. Depending on the level of organization and involvement, collective violence takes the form of one of four broad categories of behaviour: lynching, rioting, vigilantism and terrorism (Black Reference Black2004; de la Roche Reference De la Roche1996). These various outcomes exist in part because collective violence is often carried out by individuals on account of their identification with the collective and the victim’s identification with theirs rather than by collectives themselves (Gould Reference Gould1999).
In recent times, the study of collective violence has primarily focused on terrorism and the more recent, broadened conception of violent extremism (VE), which includes various sub-terroristic crimes, such as hate crimes. Criminology has played a vital role in developing a more nuanced understanding of VE (LaFree, Weerman, and Bijleveld Reference LaFree, Weerman and Bijleveld2020). This body of research has found that VE and crime broadly overlap concerning spatial and temporal patterns (e.g. Hasisi et al. Reference Hasisi, Perry, Ilan and Wolfowicz2020b; Perry Reference Perry2020), patterns of recidivism (e.g. Hasisi et al. Reference Hasisi, Carmel, Weisburd and Wolfowicz2020a), and risk and protective factors (e.g. Wolfowicz et al. Reference Wolfowicz, Litmanovitz, Weisburd and Hasisi2021), as well as concerning situational prevention (e.g. Perry et al. Reference Perry, Apel, Newman and Clarke2017) and deterrence effects (e.g. Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023).
Among the significant publications cited here, the name David Weisburd appears repeatedly. This is perhaps not surprising, given that Weisburd’s (Reference Weisburd1988) doctoral dissertation on “Jewish settler violence” (JV) in the Israeli context was one of the first criminological inquiries into collective violence and a key influence for subsequent models of collective violence (e.g. Black Reference Black1993, Reference Black2004; de la Roche Reference De la Roche1996). Weisburd (Reference Weisburd1989) found that JV was a reaction to Palestinian violence (PV) and a perceived lack of an adequate formal response to PV. Given the findings and the particular political, social and security climate in Israel, Weisburd (Reference Weisburd1988) predicted that JV would probably increase and become increasingly difficult to prevent.
The issue of JV has become a central topic both in Israel and abroad. However, despite the attention it has received, there has been a surprising lack of quantitative research on the issue. In this study, we follow Weisburd’s (Reference Weisburd1989) treatment of JV as a form of social reaction, self-help and social control and examine how it responds to PV and formal control. Our analysis draws on data from the Israel Security Agency (ISA), the Palestinian Authority, the United Nations (UN) and open sources. Through a series of generalized negative binomial (NB) regression models and Newey–West ordinary least squares (OLS) models, our analysis provides evidence that JV is a social reaction to PV and (perceived) weak formal control. The results also demonstrate that it is not only in the case of terrorism, but also other forms of collective violence, that particularly harsh formal control mechanisms (such as administrative detention) are prone to iatrogenic effects.
“Jewish Settler Violence”
Arab-on-Jewish and Jewish-on-Arab violence has been an ongoing theme since and before the creation of the modern State of Israel in 1948. However, it is the violence that has occurred in the territories captured by Israel from Jordan in the six-day war of 1967, which Israel refers to by its biblical names of Judea and Samaria (J&S) and which is commonly referred to as the West Bank (WB),Footnote 1 which has been the main focal point in recent times. This region is home to over two million Palestinians and about 500,000 Israeli Jews (both numbers are disputed), which, since the 1993 Oslo Accords between Israel and the Palestinian Liberation Organization (PLO), is split into three areas: Area A under full Palestinian control; Area B under Palestinian administrative and civil control but Israeli security control; and Area C under full Israeli control. Most Palestinians reside in Areas A and B (although some reside in C), whereas the Jewish population resides in Area C. In 2002, following a string of suicide bombings, Israel commenced construction on a large security barrier that would more or less follow the 1948 border (the “green line”) and has since completed construction on over 500 km of the planned 560 km barrier. It has been found that the barrier was successful in reducing suicide bombings. However, it led to an increase in overall terrorism in some sections adjacent to the barrier and an increase in lighter forms of violence, such as Molotov cocktail attacks throughout J&S and the WB, where most violence is now concentrated (Perry et al. Reference Perry, Apel, Newman and Clarke2017).
While Israel initially allowed some Jewish settlements that had existed before 1948 to be reconstructed immediately following the six-day war, settlement construction commenced only in the late 1970s. By the 1980s, reports of Jewish settlers on Palestinian violence (JV) increased, and the number of Palestinians killed by settlers under various circumstances reached a peak in 1994, followed by a sharp decline (Pedahzur and Perliger Reference Pedahzur and Perliger2003). Then, in the mid-2000s, the “price tag” (Tag Mehir) phenomenon emerged, with acts primarily relegated to property damage and defacement. These and other types of JV are often carried out by “hilltop youth” (No’ar HaGva’ot), a “loosely connected group of young Israelis that creates and populates many of the outposts in the West Bank” (Eiran and Krause Reference Eiran and Krause2018, 638).Footnote 2 Whilst the UN previously defined “price tag” as its phenomenon, it now forms part of a broader category of JV (Eiran and Krause Reference Eiran and Krause2018). However, the term “price tag”, which was chosen by Jewish settler activists, serves to invoke the motivations and objectives that underpin JV. Even in accounting for the more ideological, religious and nationalistic motivations, which are beyond the scope of this paper to delve into, the underlying claim remains that such violence is (1) a response to PV and the (perceived) lack of an adequate formal response, and (2) a response to perceived biases by authorities who arrest settlers and destroy settlements (Nir Reference Nir2011).
Collective Violence as Social Reaction, Self-Help and Social Control
Previous work on JV has mostly viewed it from the perspective of collective violence, which is usually but not always sub-terroristic (Gazit Reference Gazit, Bjørgo and Mareš2019; Pedahzur and Perliger Reference Pedahzur and Perliger2003). As can be inferred from the rationalizations and motivational claims for JV, Weisburd (Reference Weisburd1989) described it as a classic example of crime (and collective violence specifically) as a form of “self-help” or “the expression of a grievance by unilateral aggression such as personal violence or property destruction” (Black Reference Black1983, 34). Weisburd (Reference Weisburd1989) drew on Black’s (Reference Black1983, Reference Black1993) perspective of crime carried out in response to the actions of another as being a form of social control and crime carried out in response to the law, the enforcement response, or the lack thereof, as a form of self-help. Much crime is oriented toward redressing grievances, and especially in the case of retaliatory crime, it is used to prevent crime, especially when formal controls have failed or do not exist. In such cases, violent self-help is a form of social control aimed at controlling unwanted behaviour and (re)enforcing social norms (Black Reference Black1983).
According to this perspective, collective violence is employed to address the need for justice by members of a collective who, for whatever reason, cannot rely on the formal controls offered by the criminal justice system. In many cases, this is due to limited access to the criminal justice system, sometimes due to bias and other times due to a lack of resources, but also for potentially other reasons. In particular, when the justice system fails to deal with source-grievance crime properly, it can give rise to vigilantism. This potential social reaction is even more prominent when collective violence is source-grievance crimes (Bell Reference Bell2002; Black Reference Black1993; McDevitt et al. Reference McDevitt, Balboni, Garcia and Gu2001; Robinson Reference Robinson2015; Wexler and Marx Reference Wexler and Marx1986).
Similar to gang violence, this means that collective violence is mostly retaliatory. For example, members of group A targeted by a terrorist attack by members of group B may not have recourse to justice or retribution, either because the offender is dead or has escaped detection. Additionally, when group B’s terrorism against group A increases, this is viewed as evidence of an inadequate formal response. When punishments are infrequent and weak, the need for justice and revenge is more likely to be sought through self-help, such as vigilantism (Phillips Reference Phillips1987). This appears to be the case for hate crimes, which can serve as a form of retaliation for crimes committed by members of a target group, including hate crimes themselves, or also by the state, whose crime may be a perceived lack of response, and thereby their complicity (Lickel et al. Reference Lickel, Miller and Schmader2006; McDevitt et al. Reference McDevitt, Levin and Bennett2002). Hate crimes and vigilantism are generally less organized than other types of collective violence, such as terrorism, but display patterns in terms of timing, space and symbolism of targets (de la Roche Reference De la Roche1996, Reference De la Roche2001; Green, Strolovitch, and Wong Reference Green, Strolovitch and Wong1998; Tilly Reference Tilly2003).
Collective violence as a social reaction is most likely to emerge when the nature of the source-grievance crime(s) generates a strong identification with the victims, are perceived as representing a particular threat to the community, and, as above, when the formal response toward them generates frustration and a lack of trust among the authorities (Black Reference Black1993; Shotland and Goodstein Reference Shotland and Goodstein1984). Such violence is, therefore, dependent on strong social cohesion, differentiating it from other forms of collective violence, such as terrorism, which emerges under conditions of weak social cohesion (Gurr Reference Gurr1993). Additionally, given its self-help function, reacting to a crime that formal social control has failed to deter, such collective violence is less sensitive to deterrence by formal social controls. If anything, excessive attempts at formal control can generate a backlash effect, as it can contribute to the sense of selectivity that underpins the initial social reaction (Black Reference Black2004; Tankebe Reference Tankebe2009). Evidence for this has certainly been found concerning terrorism, with most studies on the topic finding that “hard” tactics are more likely to generate iatrogenic than deterrent effects. While traditional criminal justice approaches, such as arrests and convictions, can have a deterrent effect, long prison sentences can also generate backlash effects (Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023).Footnote 3
Figure 1(a) depicts how JV operates according to this model. Here, when the source-grievance crime (PV) is high, and formal responses against it are low, while formal responses against JV are high, JV is predicted to increase. Conversely, when PV is low, formal responses against it are high and formal responses against JV are low, so JV is predicted to decrease. Figure 1(b) simplifies this as JV increases in response to increasing PV and selective enforcement.
Figure 1. Predicted direction of Jewish settler violence (JV) according to the theoretical model. PV, Palestinian violence.
Following similar patterns, right-wing extremism and hate crimes targeting Muslims have been found to increase following Islamist terror attacks (Byers and Jones Reference Byers and Jones2007; Disha, Cavendish, and King Reference Disha, Cavendish and King2011). However, Islamist terrorism is unaffected by increases in hate crimes (Deloughery, King, and Asal Reference Deloughery, King and Asal2012; Mills, Freilich, and Chermak Reference Mills, Freilich and Chermak2017). Similarly, King and Sutton (Reference King and Sutton2013) found that while anti-Black hate crime increases in light of anti-White hate crime, there is no evidence for an effect in the opposite direction. As such, evidence suggests a unidirectional relationship in which certain types of collective violence are a social reaction to other types (Benier Reference Benier, Schweppe and Walters2016). Relatedly, Python, Brandsch, and Tskhay (Reference Python, Brandsch and Tskhay2017) point out that while terrorism is often thought to be the outcome of ethnic conflict, it also targets specific areas to create or provoke an escalation of sub-terroristic ethnic violence. In line with this hypothesized direction, Brandsch and Python (Reference Brandsch and Python2021) found that violent rioting increases in response to indiscriminate terrorist attacks, being both spatially and temporally proximate to the preceding terrorism event(s). The study found that, on average, one violent riot will occur for every 25 terrorist events. As the authors suggest, at least part of this relationship can be explained by the state response (or lack thereof) to the initial terrorism and the rioting (Brandsch and Python Reference Brandsch and Python2021). Similarly, Bell (Reference Bell2019) found that terrorist assassinations targeting officials increased spontaneous, sub-terroristic collective violence, namely social unrest, interpersonal violence and rioting. Unfortunately, none of these studies has been able to examine the full model that accounts for reaction to and in the presence of formal responses.
Returning to the case of JV, there is qualitative evidence showing it is motivated by perceptions of a lack of an effective, formal response to PV and perceived selectivity (e.g. Yassan Reference Yassan2023). However, we are only familiar with three quantitative studies. In one study, Munayyer (Reference Munayyer2012) found that increased PV is associated with a reduction in JV, while formal actions against settlements were associated with an increase. As such, Munayyer (Reference Munayyer2012) rejects the idea that JV is a reaction to PV, given that the study’s data show a 95% decrease in PV commensurate with a more than 300% increase in JV. Unfortunately, this study suffered from significant flaws, such as using daily, contemporary measures of the variables, being limited to a single year, and using dummy variables to represent government actions. Perhaps most importantly, Munayyer needs to report the source of their data for PV, and as we show below, some sources are prone to severe underreporting. In contrast, Magid (Reference Magid2020) found that between 2010 and 2015, increased clashes between Israeli security forces and Palestinians and more Jewish victims of PV increased the likelihood of JV. Eiran and Krause (Reference Eiran and Krause2018) found an apparent relationship between the destruction of Jewish settlements and “price tag” incidents.
Israel’s Response to Jewish Violence
It has repeatedly been suggested that JV has increased partly due to a biased, inadequate response from the state where JV is under-policed (Eiran and Krause Reference Eiran and Krause2018; Shalhoub-Kevorkian and David Reference Shalhoub-Kevorkian and David2016). This claim mirrors those made of biased responses toward other types of crime, and hate crime in particular, based on factors such as race or religion. However, it must first be understood that clearance rates for hate crimes are notoriously lower than non-bias crimes, partly due to most hate crimes being against property, which has a lower rate of witnesses (Armstrong Reference Armstrong2019; Lantz, Gladfelter, and Ruback Reference Lantz, Gladfelter and Barry Ruback2019; Lyons and Roberts Reference Lyons and Roberts2014). Compared to hate crimes involving interpersonal violence, hate crimes targeting property are always less likely to lead to an arrest. Only 9% of such cases in Canada led to arrests, and only 7.5% led to charges (Armstrong Reference Armstrong2019). In the UK, a similar figure of 7% of hate crimes targeting property led to an arrest (for the year 2021), while in Sweden (between 2007 and 2020), there was an average of 4%.Footnote 4 Secondly, most studies have found that offender–victim dyad identities do not have an impact on hate-crime clearance rates (Armstrong Reference Armstrong2019; Lantz et al. Reference Lantz, Gladfelter and Barry Ruback2019; Lyons and Roberts Reference Lyons and Roberts2014). As such, the clearance rate of 9% by charge for JV events between 2005 and 2019 (Yesh Din Reference Din2020) aligns with rates for analogous offences from elsewhere. Concerning potential bias, a 2021 government report on rock-throwing in Jerusalem between 2015 and 2021, a common feature of low-intensity collective violence in Israel (at least compared to other forms of violence), claimed a charge rate of 26.6% for Jews and 25.7% for Arabs (Yachimovich-Cohen Reference Yachimovich-Cohen2021).
Whilst more macro-level biases probably exist, there is evidence that the Israeli state treats JV seriously. Israel views JV as posing a particular security threat based on the view that it can further increase tensions and encourage more PV and that it is damaging to the character of the state. As such, following an uptick in JV in 2013, several high-level discussions in the Israeli parliament took place, leading to the establishment of the Nationalistic Crimes Unit (NCU), a new division in the police (Judea and Samaria division) specifically tasked with combatting nationalistic crimes. The unit began operations in March 2013, and according to reports, from its outset, it received significant resources in terms of workforce (Israel Ministry of Justice 2015). The NCU’s toolbox includes broad powers concerning administrative orders, which trace their roots to the British Mandate period. While the NCU’s creation was applauded by the UN, these measures, which have also been used extensively against Palestinians, have been criticized by the United Nations Human Rights Council (2016). The most serious type of order, administrative detention, involves incarceration without trial and often limits access to legal representation. While such orders are intended to be limited to emergency cases of “ticking timebombs” to prevent an imminent attack, they are known to have been used more liberally (Cohen-Almagor Reference Cohen-Almagor1997).
Other administrative orders used include distancing, communication and house arrest, which are more limited to Jewish suspects. Distancing orders prohibit individuals from access to certain areas for some time, ranging from specific events (such as a gay pride parade) up to prolonged prohibitions from entry to whole regions, in particular, J&S/WB, even if they reside there (Abraham Reference Abraham2014). Communication orders prohibit contact with other specified individuals and often accompany distancing or house arrest orders. House arrests can either be ordered through administrative procedures or the courts. Frequently, courts order house arrest instead of the prosecutor’s request for extended detention. While under house arrest, police station reporting or spot checks are common (The Jewish Voice 2020). In 2013–2014, 13 and 19 administrative orders were issued, respectively. However, in 2015, this increased to 45, with 90 charges laid, more than double the average from the preceding years.Footnote 5 The main methods employed by the NCU represent a set of “hard”, deterrence and incapacitation-oriented tactics. However, as discussed above, such tactics are often ineffective and may be associated with iatrogenic effects (Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023).
The Current Study
The research on how various forms of collective violence respond to formal control measures remains relatively underdeveloped. Most tests have been limited methodologically, and a relatively small number of contexts have been examined (Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023). Additionally, as the above review demonstrates, there are even fewer studies that determine if and how violence may represent a form of social control by analysing how it responds to source-grievance crimes and formal responses both to the source-grievance crimes and the response crimes (Díaz-Faes and Pereda Reference Díaz-Faes and Pereda2022).
In this study, we set out to investigate two related issues in the context of JV in Israel, a particular form of collective violence which claims to be a social reaction to Palestinian-Arab violence (PV) and a selective response by the state’s formal control. Like all contexts, the Israeli context, and that of J&S/WB in particular, is unique and may not be comparable to other settings. However, Israel has often served as a case study in research on collective violence, primarily in the case of terrorism. This is partly due to sufficient events enabling meaningful testing of theoretical propositions and hypotheses. We sought to test whether such violence follows the patterns of the social reaction/self-help framework. To do so, we first sought to test whether JV increases in response to PV and the various government responses to that phenomenon. By the theoretical framework, we hypothesized that JV would increase following more frequent incidents of PV but decrease when there was a more sufficient formal response to PV. Second, we sought to investigate how formal control measures directed at JV make an impact on it. Given the theoretical framework and existing evidence, we hypothesize that “hard” tactics will either be ineffective or potentially generate iatrogenic effects.
METHODS
Data and Variable Construction
Data for this study were derived from several different sources. As the data sources cover different periods, we restricted our observation period to the timeframe with the greatest full overlap, namely January 2009–December 2022. Beyond this being the timeframe in which the data overlap, January 2009 is also significant for marking the end of Operation Cast Lead, a key event in the Israeli–Palestinian conflict.
Most studies on governmental responses to VE rely on dummy variables, dichotomous variables like “repression” or “no repression”. They, too, tend to over-aggregate independent variables (IVs), conflating a wide array of counterterrorism measures into one variable (Chenoweth and Dugan Reference Chenoweth and Dugan2011; Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023). However, exceptions exist, such as the Government Actions in Terror Environments (GATE) database developed by Dugan and Chenoweth (Reference Dugan and Chenoweth2012), which covers several countries, including Israel. The database reports monthly counts of PV and Israeli responses, which are classed as retributive or conciliatory. However, it needs to improve its reporting, partly due to its reliance on international English-language media (Chenoweth and Dugan Reference Chenoweth and Dugan2011). We rely on more local data sources, which have been found to more accurately reflect the magnitude of incidents, both in terms of attacks and government responses (e.g. Benmelech, Berrebi, and Klor Reference Benmelech, Berrebi and Klor2015; Freedman and Klor Reference Freedman and Klor2023; Perry et al. Reference Perry, Apel, Newman and Clarke2017). Moreover, our dataset includes formal responses to JV, absent from GATE or other known sources.
JV’s official Israeli data concerning our dependent variables (DVs) are not readily available. However, since mid-2004, the Palestinian Monitoring Group (PMG) has been publishing monthly incident reports that include, among other items, JV.Footnote 6 The data report on various types of incidents, including intimidation or physical attacks, and attacks on property, including but not limited to trespassing, preventing access and property damage (see Appendix 1 for examples). The data have also been one of the primary sources for data on JV collected by the United Nations Office for the Coordination of Humanitarian Affairs (OCHA) since 2006. Unlike the PMG data, the OCHA requires each incident to be validated by at least two independent sources to be eligible for inclusion. Additionally, the OCHA data also disaggregate incidents by event type. We use both datasets to construct three different measures of our DV, two for all incidents and one for “interpersonal violence” incidents or incidents resulting in Palestinian casualties. This aims to identify whether our analysis is sensitive to differences across the datasets. Both datasets sometimes report incidents where no apparent base crime or offence under the criminal code can be readily ascertained (see Appendix 2 for examples). It is thus important to consider the potential misreporting when drawing conclusions based on these data. Nonetheless, the PMG is one of the few official sources providing information on JV, maintaining a reasonable consistency across the observed period. We do note, however, that whilst we are only aware of one study having used the JV data from the PMG (Munayyer Reference Munayyer2012), several studies have made use of the UN data (e.g. Calì and Miaari Reference Calì and Miaari2015; Haran Diman and Miodownik Reference Haran Diman and Miodownik2022; Hatz Reference Hatz2019).
In the PMG data, there was a mean of 68.49 (standard deviation [SD] = 42.08) incidents per month, whereas in the OCHA data, there was a mean of 59.29 (SD = 33.83) incidents. However, as shown above, in Figure 2, the datasets overlap considerably. For the interpersonal violence data, there was a mean of 8.58 (SD = 6.21) monthly incidents.
Figure 2. Reported monthly incidents of Jewish violence (2009–2022) by the Palestinian Monitoring Group (PMG) and the United Nations Office for the Coordination of Humanitarian Affairs (OCHA) and interpersonal violence-only incidents as reported by the OCHA. oPT, Occupied Palestinian Territory.
Concerning PV, data were derived from the monthly reports published by the Israeli Security Agency (ISA), Israel’s domestic intelligence agency, akin to the US Federal Bureau of Investigation (FBI). The data stem from the period of January 2009 until December 2022 and include all incidents that the organization classifies as “terrorism”, namely attacks involving Molotov cocktails, arson, explosive devices, firearms, bladed weapons and vehicles. The reports only occasionally include other low-intensity attacks, such as stone-throwing, when they are associated with serious injury. As such, the data do not include the thousands of annual rock-throwing incidents, which, to our knowledge, are not recorded in any readily available official data source. The ISA annual reports classify incidents as for all categories except Molotov cocktails and arson. Given this distinction and the potential for differential effects, we disaggregate PV into two levels, namely “serious” and “light”. While the ISA data cover a geographic scope encompassing the entirety of Israel’s geographical area, the overwhelming majority of incidents recorded occurred in J&S/WB and the adjacent environs of Jerusalem. As demonstrated elsewhere (see Perry et al. Reference Perry, Apel, Newman and Clarke2017), official Israeli data tend to be more comprehensive than widely used open sources, such as the Global Terrorism Database (GTD). In Figure 3, we plot our data against those of the GTD. The GTD data also include a non-comprehensive but relatively large count of rocket attacks from Gaza, making the discrepancy between the two data sources even greater. In addition, we note that the GTD data are only currently available until 2020.
Figure 3. Overall Palestinian-Arab violence (2009–2022) from the Israel Security Agency (ISA) compared to data from the Global Terrorism Database (GTD).
Regarding state responses to PV, data for three of the four variables were again derived from the PMG, namely arrests (mean = 396.35, SD = 134.76), deaths (mean = 5.77, SD = 7.20) and retributive house demolitions (mean = 0.64, SD = 1.25).Footnote 7 Data for the fourth variable, incarcerations for “security”-related offences, were retrieved from Ha’Moked, a human rights organization that receives its data through an ongoing freedom of information agreement with the Israel Prison Service (IPS). Over the period, the mean number of incarcerated “security prisoners”, those accused of security-related offences broadly connected to terrorism, was 5,545.55 (SD = 960). From 2009 to 2012, there was a sharp decline in the number of security prisoners, then an increase from 2013 to mid-2016. From this time, the numbers again decreased significantly, reaching their lowest levels in 2022Footnote 8 (see Figure 4).
Figure 4. Monthly counts (2009–2022) of Palestinian-Arab violence (serious and light), formal control responses to Palestinian-Arab violence, and formal control responses to Jewish-settler violence.
As noted above, no known official data sources are available regarding the formal control of JV. Additionally, such events are not regularly or broadly reported in mainstream media. We identified Ha’Kol Ha’Yehudi (The Jewish Voice), an independent news site associated with the settler movement, which conducts investigative reporting into issues specifically related to the state’s treatment of settlers as an appropriate open source.Footnote 9 We searched for and retrieved all items that included terms such as “arrest”, “detention” and “administrative order”, leading to hundreds of documents from which the data were subsequently extracted and aggregated into monthly counts of the different types of state actions employed to control JV. The final data, as depicted in Figure 4 and summarized in Table 1, resulted in monthly counts for arrests, distancing orders, house arrests, non-communication orders and administrative detentions.
Table 1. Descriptive Statistics for Main Variables (n = 168 Months for All Variables)
PMG, Palestinian Monitoring Group; UN, United Nations; OCHAoPT, United Nations Office for the Coordination of Humanitarian Affairs, Occupied Palestinian Territory; ISA, Israel Security Agency; N/A, not applicable; PCBS, Palestinian Central Bureau of Statistics.
We acknowledge that open sources have various limitations, especially in collective violence, including missing data, selectivity bias and misreporting. Of particular concern is that the geopolitical context of the study may exacerbate reporting bias. This should be kept in mind during the conclusion-drawing process, as open-source limitations cannot be eliminated (Chermak et al. Reference Chermak, Freilich, Parkin and Lynch2012; Cubukcu and Forst Reference Cubukcu and Forst2018; Dugan and Distler Reference Dugan, Distler, LaFree and Freilich2016; Freilich et al. Reference Freilich, Chermak, Arietti and Turner2024; LaFree Reference LaFree, Muro and Wilson2022). However, while The Jewish Voice news site is openly and knowingly affiliated with the settler movement and a right-wing political agenda, potentially raising concerns about reporting bias, many of the articles included photographic images of the orders, enabling us to confirm the authenticity of the claims made in the various articles. Additionally, we were able to identify some official data for cross-referencing purposes. For example, according to a report by the Knesset Research and Information Center (2015), there were four Jewish administrative detainees as of November 2015, which overlaps with our data. So, too, ISA annual reports for 2012, 2014 and 2015 included data on the number of administrative orders and criminal charges issued at the aggregate level. Through these data, we confirmed a broad overlap with our data and established that, if anything, our data suffer from some underreporting, rather than the overreporting that would be expected in the presence of ideological bias.Footnote 10 For administrative detention and distancing orders, it was also possible to identify corroborating media reporting from other right-leaning outlets, such as Israel National News. Unfortunately, we could not identify additional sources for cross-referencing purposes for arrests, house arrests and non-communication orders. As with the data on PV, whilst we acknowledge the limitations of these data and the potentially biased nature of their source, there is sufficient evidence to suggest that they possess an acceptable level of reliability.
Another data challenge was acquiring population statistics, as official Israeli sources have been limited. Jewish population statistics for J&S/WB were obtained from B’tselem, a left-wing non-governmental organization that monitors settlement activities and whose data have been used widely in research (Benmelech et al. Reference Benmelech, Berrebi and Klor2015; Freedman and Klor Reference Freedman and Klor2023; Hatz Reference Hatz2020). Palestinian population data were derived from the Palestinian Central Bureau of Statistics.Footnote 11 Population size has consistently been found to be a key, positive predictor of collective violence (Gassebner and Luechinger Reference Gassebner and Luechinger2011; Jetter and Stadelmann Reference Jetter and Stadelmann2019) and negatively correlated with social cohesion, a predictor of collective violence (Haran Diman and Miodownik Reference Haran Diman and Miodownik2022).
Analytic Strategy
As opposed to the analytic approaches taken in analysing various forms of high-volume crime, previous work on deterrence and collective violence has recognized the relatively low counts of both the DVs and IVs and their skewness as key methodological considerations. Concerning the DV, it has been shown that results can differ considerably when using counts or event rates, such as the number of events per 100,000 inhabitants (Jetter and Stadelmann Reference Jetter and Stadelmann2019; Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023). Additionally, regarding IVs, there are concerns that using the number of arrests divided by the number of events may produce negative correlations that are artefacts of the numerator of the DV serving as the denominator of the IV (Pogarsky and Loughran Reference Pogarsky and Loughran2016). Using counts or inverse-hyperbolic sine (IHS)-transformed counts of such predictors may be a preferred solution (Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023).Footnote 12 We follow prior research by employing a series of count-based regression. Due to overdispersion in our data, we employ NB regression (Weisburd et al. Reference Weisburd, Wilson, Wooditch, Britt, Weisburd, Wilson, Wooditch and Britt2022). While these models account for overdispersion, it is also necessary to account for autocorrelation. As such, we specify a set of generalized linear models (GLMs) with an NB estimator, which also allows us to compute Newey–West standard errors, which are robust to heteroskedasticity and autocorrelation (Newey and West Reference Newey and West1987). To check the robustness of our results to different specifications, we also conducted a Newey–West OLS regression on the IHS transformed variables, with the DV being the rate of JV per 100,000 settler inhabitants and all predictors being the IHS transformed counts (see Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023). Cumby–Huizinga tests were used to identify the lag numbers for the different sources of the DV for the specification of the Newey–West standard errors (Baum and Schaffer Reference Baum and Schaffer2015).
Given the nature of the inquiry and that some of the predictors are weakly endogenous, all predictors were entered as lagged variables (one month), including a lag of the DV regressed on itself.Footnote 13 In addition, it is common in such models for seasonality to be addressed through the use of monthly dummy variables (Cameron and Trivedi Reference Cameron and Trivedi2013). However, some have pointed out that this leads to a more deterministic model, especially when a lagged DV is included, and that it can also lead to issues (Wilkins Reference Wilkins2018). While others prefer to address this issue by detrending the data using a first-differencing approach, this approach also carries its own set of problems (see O’Brien Reference O’Brien1996). However, given that Dickey–Fuller tests showed that we could reject the presence of a unit root in all variables (p < 0.001), we carried out three sets of models. First, we run a set of models without using monthly dummy variables. Then, we run a second set of models in which they are included. Following this, we run a set of models in which we drop the dummy variables and replace them with a set of sine–cosine pair terms (Cox Reference Cox2006), an econometric approach that has been used with time-series data for crime (e.g. Enders, Pecorino, and Souto Reference Enders, Pecorino and Souto2019; Lee, Pecorino, and Souto Reference Lee, Pecorino and Souto2023; Wheeler and Haberman Reference Wheeler and Haberman2018). The inclusion of these terms is equivalent in their implications to monthly dummy variables, although they are not to be interpreted (Ghysels and Osborn Reference Ghysels and Osborn2001, 20–4). We assess differences in model fit with the Akaike information criterion (AIC) and Bayesian information criterion (BIC).Footnote 14
As discussed above, it is also important to control for population size. However, given that we are interested in investigating inter-group crime events, we actually have two populations whose sizes may be important. Rather than simply controlling for the (log) population size of either or both populations, we follow the approach of Hipp, Tita, and Boggess (Reference Hipp, Tita and Boggess2009) by calculating the conditional probability of between-group interaction between groups A and B (i ab) based on both population sizes according to Equation 1 below. Here, N A is the size of the settler population, N B is the size of the Palestinian population and N is the size of the whole population.
$${i_{ab}} = \left[ {\left( {{N_A}} \right)\left( {{N_B}} \right)} \right]/\left[ {\left( N \right)\left( {N - 1} \right)} \right]$$
RESULTS
Table 2 displays the results of the analysis. Models I–IIIa report the results for models using the PMG data, followed by models IV–VIa for the UN data and models VII–IX for the interpersonal violence-only outcome derived from the UN data. In all cases, the first models have no controls for months (models Ia, IIIa and V1a), the second includes the monthly fixed effects (models IIa, IVa and VIIIa) and the third includes the sine–cosine pair terms (models IIIa, VIa and IXa). The findings were consistent across all models and the three DVs, at least substantively. In terms of model fit, models that included monthly dummy variables demonstrated the best fit, although those using the sine–cosine pair terms also demonstrated an improvement over those without.
Table 2. Generalized Negative-Binomial Regressions on Jewish Violence for 2009–2022 (n = 168 Months) a
a Coefficients reported with Newey–West standard errors. All predictors were lagged (t – 1).
***p < 0.001, **p < 0.01, *p < 0.05, †marginally significant (p < 0.1).
With respect to the impact of the lagged DV, in the PMG data, each additional incidence of JV in the preceding month increased JV in the following month by 0.3%, whereas in the UN data, the increase was between 0.4 and 0.6% depending on the model. In the UN data for interpersonal violence, the increase ranged from 2.3 to 2.9% depending on the model. Concerning how JV responds to serious incidents of PV, in the PMG data, a one-unit increase was associated with approximately a 1.0% increase, whereas in the UN data, it was associated with a 0.5–0.6% increase, depending on the model. Depending on the model, the increase ranged between 0.7 and 0.9% for interpersonal violence in the UN data. Conversely, light forms of PV were found to have a negative association with JV. In all datasets, a one-unit increase was associated with a 0.2% decrease in JV, with a maximum of a 0.3% decrease in the case of interpersonal violence in the UN data.
Concerning formal controls against PV, the arrest estimate consistently remained well below statistical significance, except in the case of interpersonal violence in the UN data, where it was marginally significant (p < 0.10) and had a small, positive sign. On the other hand, with respect to the number of security prisoners, the estimate was consistently significant, albeit with exceptionally small coefficients indicating between a 0.01 to 0.02% decrease in JV for a one-unit increase in the prisoner population. However, the issue of scale must be taken into consideration when appreciating this statistic, as an increase of 100 prisoners, for example, could make a meaningful impact on JV. In the case of killings, estimates were not consistent across DVs, remaining below reaching statistical significance in the PMG data, reaching significance in the UN data, and being marginally significant in the interpersonal violence data. In the UN data, a one-unit increase in killings was associated with a 1.1–1.3% decrease in JV, whereas the marginally significant estimates for interpersonal JV indicated a 1% reduction. Concerning home demolitions, estimates were almost always statistically significant, except in the case of model Va (UN data with monthly fixed effects), and a marginally significant estimate in model IIIa (interpersonal violence with monthly fixed effects). In the PMG data, a one-unit increase in home demolitions was associated with a 6.4–7.3% decrease in JV, whereas in the UN data, there was a 4.7–4.8% decrease and for interpersonal violence, an 8.1–8.5% decrease.
Concerning formal controls against JV, in the case of arrests, the results were at least substantively consistent across specifications. Only in model Va was the estimate not statistically significant. In the PMG data, a one-unit increase in arrests was associated with a reduction in JV of 0.8–0.9%, whereas the reduction was 0.5% in the UN data and 0.7–0.9% for interpersonal JV in the UN data. For house arrests, the results were less consistent, being only marginally significant in the PMG data, where a one-unit increase was associated with a 7.2–7.4% decrease in JV, statistically significant (or marginally significant) in the UN data with an associated 6.3–6.5% reduction, and non-significant concerning interpersonal JV. For both distancing and non-communication orders, across all models, the estimates remained well below the threshold of statistical significance.
Regarding administrative detention, estimates were statistically significant across all models except for model Va, where the estimate was only marginally significant. In the PMG data, a one-unit increase in administrative detentions was associated with a 31.3–37.4% increase in JV, and estimates were of a similar magnitude in data for interpersonal JV with a 26.4–33.8% increase. In the UN data, estimates were smaller but still salient effects with increases between 13.6 and 14.3%.
Robustness Check
As mentioned above, we re-ran the analysis using the IHS-transformed number of events divided by 100,000 of the settler population as the DV and IHS-transformed counts of all IVs using Newey–West OLS regression (Table 3). The results were substantively consistent with our main results to a significant degree. Following the order of our main results, the first model for each DV included no time controls, the second model included monthly dummies, and the third model included Fourier terms.
Table 3. Newey–West Ordinary Least Squares Regressions on Jewish Violence for 2009–2022 (n = 168 Months) a
a Coefficients reported with Newey–West standard errors. All predictors were lagged (t – 1).
***p < 0.001, **p < 0.01, *p < 0.05, †marginally significant (p < 0.1).
While serious PV was statistically significant in the PMG and OHCR data, it was not significant in the interpersonal violence data. Interestingly, the light PV estimate was below the level of statistical significance across all DVs and models. Concerning formal responses against PV, the effects of arrests, prisoners and killings were consistent with the main results. However, while the estimates for home demolitions were mostly consistent, they were occasionally only marginally significant and below statistical significance in the case of model Vb. Concerning formal responses against JV, arrests were only significant in two of the models examining interpersonal violence (models VIIIb and IXb). The findings for distancing, communication orders and house arrests followed the primary analysis’s findings. Concerning administrative detention, in models IVb and VIb, the estimate was only marginally significant, whereas in model VIIb, it was below the level of statistical significance.
DISCUSSION
Our study sought to identify whether JV, a specific form of collective violence, demonstrates patterns consistent with the theoretical framework that positions such violence as a form of crime as social control. Under this model, collective violence by group A is a response to collective violence by group B targeting group A and a (perceived) biased formal response toward the groups. In line with this, we hypothesized that JV would increase as PV increases, in conjunction with a less-than-adequate state response. Additionally, we also hypothesized that harsh formal responses would be likely to engender iatrogenic effects. We tested our hypotheses across various specifications, which displayed substantively similar findings.
Our results provide evidence in support of the social reaction model and our hypotheses. That is, we consistently found that JV increases in response to PV, similar to how hate crime has been found to increase following terrorism. However, in this regard, we note that JV is only sensitive to serious incidents of PV, such as shootings, stabbings and vehicular attacks. In fact, we found that the more frequent acts of low-intensity violence in the form of Molotov cocktails, which normally target drivers of vehicles in motion, were generally associated with reductions in JV. First, this finding points to the potential importance of disaggregating the types of violent incidents. Perhaps failure to do so could explain the opposing findings of Munayyer (Reference Munayyer2012), who found that JV is reduced by PV. Second, serious attacks are far more likely to lead to serious injury or death, which are likely to cause more anger and for which citizens expect a more significant formal response. Conversely, frequent Molotov cocktail attacks rarely lead to serious injury, and perhaps citizens have become accustomed to both their occurrence and the lack of formal response.
While the OLS models did not find an effect for low-intensity PV, the results from the primary analysis may be understood in light of research on the psychological impacts of collective violence. Research from Northern Ireland has found that different types of terrorism have differential impacts on citizens’ psychological well-being. Here, only terrorism resulting in fatalities was found to have an impact on well-being significantly. Among the various explanations suggested is that certain types of low-intensity incidents are unlikely to be widely reported, and citizens may be unaware of their occurrence and frequency. Additional findings supported this proposition that incidents only negatively make an impact on well-being when they occur close to the respondents’ residences, increasing the incidents’ salience (Bryson and MacKerron Reference Bryson and MacKerron2018). In addition, recent evidence from Israel shows that Israelis’ punitive attitudes differ considerably across various types of violence, with less punitive attitudes associated with lower-intensity events such as stone-throwing and higher for more serious, lethal events such as shootings (Levy and Rozmann Reference Levy and Rozmann2023).
The social reaction model posits that it is not just the occurrence of a source-grievance crime that increases the likelihood of crime as social control but also the quality of the formal response to the source-grievance crime. While our analysis found that greater levels of government response to PV are associated with reductions in the likelihood of JV, this effect is limited to imprisonment and home demolitions. In the case of the former, imprisonment has declined over the period, which probably explains at least some of the variance in JV. The larger effect for home demolitions is likely to reflect the relatively low frequency of this measure compared to incarcerations, and the scale issue must be considered. Additionally, unlike frequent arrests, home demolitions are usually widely publicized in the local media, partly due to the state attempting to demonstrate the seriousness with which it deals with PV. As such, the Jewish community is more likely to be aware of such actions. Additionally, the visibility of home demolitions may satisfy community members’ demand for a salient response. An alternative possibility is that home demolitions reduce PV (Benmelech et al. Reference Benmelech, Berrebi and Klor2015), which is associated with a reduction in JV.
Regarding formal control directed against JV, we found that arrests were associated with reductions. This is in line with previous evidence that arrests are associated with deterrence in the context of terrorism (Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023). However, other tactics, specifically administrative distancing, non-communication orders and house arrests, were found to have no appreciable impact on JV in either direction. In this regard, in discussing the use and effectiveness of similar orders in countries such as France, Germany, Canada, the USA and the UK, Mehra, Wentworth, and van Ginkel (Reference Mehra, Wentworth and van Ginkel2021) highlight the importance of differentiating between short- and long-term effectiveness and policy objectives. In this regard, even if such orders do not reduce offending, they serve a broader function as symbolic tools for appeasing the population and demonstrating the state’s capacity.
Additionally, using such orders, even when short- or mid-term effectiveness cannot be established, may safeguard authorities’ ability to maintain the authority to use such measures. When considering long-term effectiveness, these considerations must be balanced against the risk that such measures can contribute to the grievances that may underpin motivation toward violence. In this regard, while some studies have found that “harsh” policies, such as crackdowns, can deter various forms of collective violence (e.g. Yang and Jen Reference Yang and Jen2018), most have found that there are either no effects or iatrogenic effects (Wolfowicz et al. Reference Wolfowicz, Maria Campedelli, Seaward and Gill2023). Our findings that administrative detention orders were consistently associated with iatrogenic effects provide additional evidence.
Prior research has shown that in the case of collective violence, and vigilantism in particular, it is not merely the harshness of the formal response that may be the issue but also the perceived selectivity of the response. That is, it may be that it is not only the presence or magnitude of state intervention and formal control but also its quality and character that determines its impact on violence as a form of self-help. Here, state responses lacking procedural justice and fairness may not only diminish potential deterrent effects but also add to grievances that underpin the very violence they seek to reduce (Tankebe Reference Tankebe2009). These effects are amplified in the presence of (perceived) police corruption, in particular when the police are viewed as collaborating with the criminal elements responsible for the crimes toward which the social reaction is occurring (Asif Reference Asif2023; Asif and Weenink Reference Asif and Weenink2022; Tankebe Reference Tankebe2009). These perspectives offer at least one possible explanation for our findings.
Given the nature of vigilantism, especially when it is more repeated and organized (as opposed to isolated incidents), it is important for police authorities to effectively communicate to the community about the efforts they are investing in to prevent the source crime and, if necessary, to work to develop new strategies. It is also important for authorities to build trust and communicate the risks associated with vigilante actions directly to the affected communities. Such efforts may also reduce popular support for vigilantism, which, as noted above, is necessary for it to flourish (Silke Reference Silke2001). Understanding how to prevent collective violence involves analysing both the processes of instigation or provocation and inhibition or repression. Here, inhibition or repression should not necessarily be limited to, or even primarily seen through the prism of formal control but rather through the shifting of the normative belief that collective violence is an “inappropriate, ineffective, or self-defeating mode of action” (Myers and Oliver Reference Myers and Oliver2008, 174). Policing authorities must accept that if policing source-grievance crime is inadequate, vigilantism is very likely to occur (Silke Reference Silke2001).
While primarily included as a control variable, the identified effects for population size also deserve attention as they have potentially important policy implications. This is partly due to Weisburd’s (Reference Weisburd1989) prediction that JV would increase as the population increased. This prediction broadly conforms with the finding that population size is one of the most important predictors of terrorism. However, concerning vigilantism, which is sensitive to social cohesion, increased population size would decrease its likelihood. Additionally, while previous research has often hypothesized that increases in the settler population exacerbate PV, they have often found this not to be the case. Indeed, we found that as the ratio of settlers to Palestinians increases, thereby increasing the probability of interaction, there is a reduction in the magnitude of JV. Policymakers should consider the role of social cohesion in developing approaches for combatting various crime phenomena, and the current case is no different.
Another issue that arises is the potential danger in labels such as “settler violence” or forcing the label of “violent extremism” or, even worse, “terrorism” on a type of violent behaviour that may be better classified under a different category. Calls to label all JV as terrorism have been made for some time (Byman and Sachs Reference Byman and Sachs2012). However, such labelling risks promoting the creation or strengthening of the very collectivist, group-based identities that increase the risk of collective violence, which such approaches aim to combat. Even beyond direct labelling, using counterterrorism resources to deal with various forms of activism, vigilantism and sub-terroristic forms of crime leads to inherent labelling. This type of stigmatization has been found to increase radicalization and extremism across a wide variety of contexts (Appleby Reference Appleby2010; Van den Broek Reference Van den Broek and Cromer2017). Labelling prisoners as terrorists, in particular, which is inherent in administrative detention, carries added risks of creating terrorist identities where they may not have existed before and adds to the grievances against the state, such as claims of selectivity and bias (Thompson Reference Thompson2020).
Limitations
Like most studies, ours is not without its limitations. First, we need to note that our approach does not seek to equate, let alone compare, Jewish and Arab violence, which are qualitatively different, including in terms of frequency, intensity and incident types. Rather, we limited our scope to examining the interplay and cycle of violence between these two phenomena and mediated by the state. In saying this, the current context examined in this study, like all contexts, is unique, and we caution against it being used to make generalizations.
Moreover, our data were aggregated temporally at the monthly level and spatially at the national level. While both PV and JV are likely to be sensitive to each other across the region due to high levels of connectedness and the relatively small size of the region, examining more localized effects would still be useful. Using weekly or even daily counts of events may uncover more nuanced effects. Additionally, both PV and JV are heterogeneous regarding the characteristics of incidents. Even in our data, the majority of incidents of JV relate to property damage rather than interpersonal violence. Whilst we examined a subset of data limited to interpersonal violence only, future research may identify differential effects when examining the phenomenon at various levels of disaggregation.
Moreover, as noted above, different data sources that we relied on in this study suffer from various biases that may lead to under- and overestimating the counts of different variables. In this regard, our measure of PV is limited to those incident types classified by the ISA as terrorism. However, there are thousands of incidents of rock-throwing, rioting, burning tyres used as roadblocks, and other low-intensity violence that occur annually. These are actually the types of incidents that Jews are most likely to fall victim to, and, hence, it is possible that their inclusion could alter the observed effects of our study. So, too, concerning JV, our data sources include many events that some may take issue with being classified as violence, and in many cases, it is questionable whether a criminal offence has even occurred. Furthermore, politically motivated and biased organizations, whether the PMG or Ha’Kol Ha’Yehudi, may overestimate counts of arrests and other factors in order to promote their narratives. Whilst we nevertheless believe that these data have offered an acceptable level of reliability, again, future research should seek to identify new data sources that enable the inclusion of the most important events and variables at a high level of reliability.
Lastly, we acknowledge that some high-profile events occurred in 2023 following the data collection and analysis carried out in this study. Perhaps the most well-known is an incident called the “Hawara rampage”, which took place in the town of Hawara (south of Nablus) on 26 February 2023. The incident involved hundreds of settlers arriving in the town in retribution for an Arab attack that killed two Jews earlier that day in the same location and carrying out violence against property and persons. Reports indicate that one Palestinian was killed, 100 injured (four seriously), and millions of shekels in property damage incurred stemming from arson. Additionally, since this event, an unprecedented number of Jews have been placed in administrative detention, with seven as of August 2023. Certainly, it will be important for new analyses to be conducted after a sufficient amount of time has passed to include and assess whether the inclusion of these events leads to any changes in the observed effects from our analysis.
CONCLUSIONS
The study demonstrates broadly the usefulness of traditional criminological frameworks, both theoretical and methodological, for examining various forms of collective violence. Whereas most applications have focused on terrorism, both conceptually and empirically, the case of settler violence is more accurately a representation of vigilantism. Despite this, the direction of the effects observed in our results broadly overlaps with those found in previous studies on various forms of collective violence, namely VE, hate crime and terrorism. As has been pointed out in some of this prior research, common wisdom about the causes of collective violence may actually reflect its effects, and our study provides additional evidence for such a possibility, highlighting the need for future research to consider such dynamics.
Our study also broadly overlaps existing research, finding that harsh measures may generate backlash rather than deterrent effects. However, it also finds that ordinary incapacitation measures can lead to small but potentially meaningful reductions. Consequently, the findings support the use of conventional anti-crime measures to curb vigilantism/Jewish violence, although not the use of anti-terrorism-oriented measures. Again, this finding and the natural conclusions to be drawn from it are not unique to the Israeli context and, as such, provide additional evidence for more balanced and evidence-based approaches to dealing with collective violence.
Acknowledgements
We would like to thank Professor Weisburd, who is both the inspiration for this study and to whom it is dedicated, for his valuable insights and comments. We would also like to thank Dr Tal Jonathan Zamir for her helpful comments and suggestions on earlier iterations of the study.
Appendix 1. Jewish Violence: Examples of Reported Offences in Palestinian Monitoring Group Reports
Appendix 2. Examples of Entries in Palestinian Monitoring Group Reports with Questionable Base Crimes
Michael Wolfowicz is senior lecturer in the Institute of Criminology, Faculty of Law, Hebrew University of Jerusalem. He completed his PhD at the Hebrew University of Jerusalem. He was a Rothschild Foundation postdoctoral fellowship recipient and conducted his postdoctoral research at University College London. His research primarily focuses on criminological applications to the study of radicalization, extremism and terrorism, both theoretically and methodologically. His work to date has included the study of terrorism and place (crime and place), recidivism, policing of terrorism, deterrence, echo chambers, social media surveillance, media effects, and risk and protective factors. He primarily makes use of quantitative methods, including meta-analysis. His broader research interests include deviant cognitions and media effects, with a focus on the impacts of network structure characteristics.
Esther Salama is a doctoral candidate at the Institute of Criminology, Faculty of Law, Hebrew University of Jerusalem. She completed her master’s degree in European and international politics at the University of Edinburgh. Her work focuses on the impacts of deterrence and incapacitation on terrorism, including the analysis of both preventative and iatrogenic effects. Her research methodologies predominantly involve quantitative approaches, such as meta-analysis and time-series analysis.
Open access
