• Non ci sono risultati.

Data collection and data cleaning

CHAPTER 3 METHODOLOGY

3.3 Data sources: police reports and judicial documents

3.3.2 Data collection and data cleaning

for the other gangs mentioned thus far, the investigation revealed that the Trinitarios had a fair level of internal organization, with leadership roles, payment of dues and regular meetings. Criminal activities constituted an essential part of the gang life and were aimed at both the financial sustainment of the group (e.g., drug-related crimes) and at the control of the territory against rival gangs (e.g., attempted murder).

• Operation Barrio 18

Operation Barrio 18 is the last and most recent major police operation conducted against Latin gangs in Milan. The investigation revealed that the Salvadorian gang of Barrio18 had a fair organizational structure (with leaders, regular meetings, and payment of weekly dues) and was involved in violent and street-level crimes, including attempted murder, aggravated robbery and drug dealing. In line with all previous gangs described, the Barrio18 was highly involved in between gang fights, mostly assaults towards the MS13.

who appeared in the intercepted conversations. More specifically, the dataset was built including general information on gang members (e.g., name, family name, nickname, country of birth, gender, date of birth, gang affiliation, role within the gang) and other individuals identified. Each person was assigned an alphanumeric ID, which allowed to prevent his/her identification and ultimately served to link each individual to the conversations dataset and co-offending dataset. The conversations dataset contains data on wiretapped conversations, including the date of the telephone call (or SMS, date of conversation in case of wiretap data obtained through spy bugs or other means), the number and identity of the participants (reported using the same ID of the attributes dataset), the gang membership and gang role of participants, and the type of activities discussed. The co-offending dataset contains data on crimes committed by more than one person and includes information related to the date of crime, 51 place, type of crime, and number and identity of offenders and victims (reported using the same ID of the attributes dataset).52

The three datasets initially built for each case study were eventually merged into three final datasets which contain relevant information extracted from all the selected case studies on individuals’ attributes, activities discussed in wiretapped conversations and co-offending events. The merge allowed to correct mistakes, inconsistencies, and to properly identify some individuals (e.g. some were only mentioned superficially in one case study while they were charged in another investigation). Ultimately, this procedure has allowed to have a better comprehension of the nature, complexity, and reliability of the data (on this aspect, see also Rostami and Mondani 2015). Table 4 shows the type of variables included in each final dataset.

51 The majority of the crimes committed by gang members or individuals offending with gang members were co-offending crimes. There were few occasions of crimes committed by solo offenders. For the purpose of this research, such crimes were excluded from the analysis.

52 More information about the data cleaning process are provided in the subsection “Data cleaning:

individual attributes, conversations and co-offending dataset” in APPENDIX A.

Table 4. Individual, co-offending, and conversations dataset and related variables

Dataset Data

Individual attributes

Alphanumeric ID

N. of police operations in which the ID appears

Police operation (e.g. Amor de Rey)

Name, family name, nickname

Sex, date of birth, country of birth

Kinship or relationship ties with other ID(s)

Information on charge for unlawful association to commit a crime (art. 416)

Information on whether the charge was handled by the juvenile justice system

Information on whether the ID was involved in co-offending

For offenders, information on offending (retrieved from co-offending dataset: N. of co-off. crimes, data first/last offense, N. of violent/property/weapons & drugs crimes)

Information on victimization

Information on criminal history

Gang-related information (membership, subgroup, role)

Information on type of activities discussed/perpetrated (retrieved from conversations dataset)

Additional information (e.g. residence status)

Conversations

ID of each segment of conversation (i.e. category) and ID of each conversation (see section 3.5.2)

Police operation of intercepted conversation

Date and time of conversation

ID of each participant in the conversation (retrieved from individual attributes dataset)

Gang membership and gang role if each participant (retrieved from individual attributes dataset)

Gang conversation (e.g., MS13, or between gang if the participants have different membership)

Type of activities discussed (coded as theme, subtheme, and category, see section 3.5.2)

Co-offending

ID of each event

Date and place of crime

Crime type (property, drugs, weapons, violent, murder/att.

murder; see Table 22 in APPENDIX A)

ID of each co-offender (retrieved from individual attributes dataset)

N. of offenders and victims

Mean age of offenders and victims

Sex and role of offenders and victims

N. of different gangs involved

Information on whether the crime was a between gang fight

The final individual attributes dataset contains information on 632 people, of whom 50% (n = 317) were identified as belonging to Latin gangs. As for the conversations dataset, the final number of unique individuals identified accounted to 364, for a total of 1,311 conversations (for further details on this dataset, see 3.5.2). As for the co-offending dataset, 232 individuals identified in the individuals dataset participated in co-offending events.53 Co-offenders belonging to less relevant gangs (or non-gang co-offenders) were subsequently excluded to analyze only members of one of the nine major gangs for which data was available as they were the primary target of the investigations. The final number of relevant gang co-offenders used for the analyses account to 200 individuals, involved in a total of 142 co-offending events.54 Table 5 summarizes the main socio-demographic and crime-related variables of gang co-offenders.

Table 5. Subset of gang co-offenders used for the analyses (n = 200)

Variable % Category and description

Sex 95.5 Male

Ethnicity

60.5 37.5 2.0

South American Central American Other (i.e. Europe, Asia)

Decade of birth

61.0 33.5 5.5

1990s 1980s 1970s

Criminal history 29.5 Yes (convicted/arrested)

Leadership role 34.0 Yes

(Leader, Underboss, High-ranking official) Juvenile at first offense 28.0 Yes (14-17 years old)

Crimes with gang

members sharing a tie 11.0 Yes

(blood, kinship, or relationship tie) Between gang crimes 73.5 Yes

(crime against a rival gang) Crimes with other gangs 29.5 Yes

(crime committed with members of other gangs) Murder/attempted murder 53.0 Yes

The majority of offenders were males (95.5%), with a prevalence of South Americans (60.5%) over Central Americans or citizens of other countries. More than half gang

co-53 This figure does not include 36 individuals who were charged for being part of an unlawful association to commit a crime but were nonetheless never involved in co-offending crimes.

54 To perform the analysis at individual and group (i.e. gang) level, information from the individual attributes and co-offending dataset were merged, to link crime data with socio-demographic and gang-related data for each offender.

offenders were born in the 1990s (61%), and nearly a third had a criminal history (29.5%).

Members with leadership roles accounted for the 34% of co-offenders and nearly a third of the sample committed the first recorded co-offense while still a juvenile (28%) and in alliance with members of other gangs (29.5%). Interestingly, a small yet relevant proportion of offenders participated in crimes with members with shared kinship (or relationship) tie (11%), while half of the offenders were involved in murder/attempted murder (53%). Co-offending events were predominantly the result of between gang feuds and this is highlighted by the fact that 73.5% of offenders participated in at least a crime against rival gang members. Lastly, more than half of gang co-offenders (53%) were involved in at least one murder or attempted murder.

The conceptualization of the variables related to individuals’ gang membership and role contained in the individual attributes dataset are discussed in the following subsection.

CONCEPTUALIZATION OF GANG MEMBERSHIP AND GANG ROLE

The identification of gang membership has long been debated in the literature, with a particular focus on measures (e.g. self-nomination) to properly identify gang members (see Esbensen et al. 2001). In this research, gang membership was assigned to individuals based on the information contained in each case study and in line with the adopted Eurogang definition (see section 1.1). Gang membership therefore was not restricted to those individuals investigated or charged for being part of a criminal gang (i.e. in legal terms, an unlawful association to commit a crime), as the police did not always gather enough evidence to charge all suspected individuals. The identification of gang membership also relied on accounts of informants, surveillance activities and/or knowledge from previous investigations, or statements of individuals (or made by rival gang members) on their affiliation made either to the prosecutor during interrogations or emerging from intercepted conversations. The combination of different sources allowed to accurately identify gang members regardless of the law enforcement perspective, thus reducing the potential bias of adopting the criminal charge as the only measure of gang

membership.55 This approach ultimately permits to have a broader picture of the gangs present in Milan in a given period, although some gangs emerge only marginally.

With regard to gang members’ role, it is the only variable of the individual attributes dataset which required an extensive data cleaning process after the data entry. The reason is twofold: first, there is the role identified by prosecutors based on the criminal code;

second, there is the internal role gang members held within their group which does not necessarily correspond to the one they were prosecuted for. On the one hand, in fact, based on the evidence collected, the prosecutors indicted individuals for being either leaders/promoter or participants of the unlawful association, as per the Italian Criminal Code (see section 3.3.1). On the other hand, most Latin gangs have internal roles and tasks assigned to members who have proved their criminal skills and loyalty to the gang.

Clearly, these gang internal roles do not precisely overlap with the label of leader or participant reported in the criminal code. Most Latin gangs had subdivision of tasks for individuals with leadership status; few others only had leaders and underbosses leadership roles, thus a less complex hierarchical structure (e.g., as the case of the Neta and the Trebol). Detailed information on gang roles emerged in intercepted conversations in which individuals spoke freely about their tasks. For these reason, the investigators were often capable to also reconstruct the gang internal structure with specific roles and tasks assigned to each individual. This ultimately resulted in having two types of variables related to the role held by gang members: Formal_role and Gang_role. The former variable refers to the role with whom gang members were charged as part of the unlawful association (“Leader/Promoter” or “Participant”). The latter variable is the role gang members held within their group and emerged through intercepted conversations or described in the manifesto (i.e. gang written rules) seized by the police. Different gang roles were grouped into four categories, three related to leadership roles and one to individuals with no specific tasks:

Leader: management of the gang, organization of meeting, assignment of gang roles and tasks, instructions and orders to gang members;

Underboss: advisor of the leader, organization and management of the gang;

55 For instance, of the 75 individuals charged in Operation Amor de Rey, only 63 were actually charged for unlawful association to commit a crime. The inclusion of additional information permitted to identify a total of 142 gang members, some belonging to gangs not even directly tackled by this operation.

High-ranking official: advisor of the leader, role subordinated to the leader and underboss but high rank that allows to give orders and control affiliates. Include duties related to the collection and management of weekly dues, the management of the gang archives and teaching of the rules to affiliates;

Affiliate: no specific duties and tasks assigned (i.e. footsoldiers, mostly new members).

A variable named Role_CLEAN was built to store the final values of gang members’

roles, result of the data cleaning and transformation process. In this regard, the individual’s gang role was preferred to the role identified by the judicial authorities to prosecute gang members. For instance, if a gang member was prosecuted as a “Leader”

but in fact held an “Underboss” role within the gang, the role for that member was coded as “Underboss”. In other cases, gang members were prosecuted as “Participant”—

possibly because not enough evidence was found to prosecute them as “Leader”—though they had a gang leadership role (e.g. high-ranking official), which was properly coded.

Moreover, individuals with no gang roles and/or prosecuted as “Participant” were coded as affiliates. Finally, individuals who were not charged but who were identified as being gang members with either leadership or affiliate roles were assigned that role (for an example of data cleaning, see Table 17 in APPENDIX A). The validity of this coding procedure was further confirmed by the final comparison between Formal_role and Role_CLEAN, which revealed that all individuals charged as “Leader” held a leadership role within the gang. More importantly, the comparison highlighted that ~80% of individuals charged as “Participant” were indeed “Affiliate”, though the remaining ~20%

had an actual gang leadership role which was properly accounted with the strategy adopted. Similarly, a total of 163 individuals were identified as gang members but not formally charged, though ~25% emerged as having a leadership role (see Table 18 in APPENDIX A).