METHODOLOGICAL ASPECTS OF GANG MEMBERSHIP: THE CASE OF THE CZECH REPUBLIC*1

Research into gangs has been primarily concentrated in the USA for many years. The important step leading to a more intensive analysis of this phenomenon in Europe has been made by creating a uniform gang definition of Eurogang research group, which has been operationalized and integrated in two waves of the International Self-Report Delinquency Study (ISRD-2 and ISRD-3). The objective of this paper is to introduce methodological issues of the definition and offer possibilities for its revision. Data suggest that each type of definition used in our study may generate a different group of respondents. Although the multivariate analysis of factors influencing gang membership demonstrates some intersections, delinquency rates among gang members defined using different definitions vary considerably.


Introduction
While research into gangs has had quite a long tradition in the United States, in Europe we began studying them only much later, although it is precisely the group nature that constitutes one of the most important features of juvenile delinquency (Warr 2002;Matoušek, Kroftová 2003).One possible reason for this reluctance might be the so-called Eurogang paradox, i.e. the belief that American gangs are represented by highly organized units involved in extremely severe crime.On the contrary, what can be considered as a paradox is the fact that "real" gangs in America only rarely correspond to that stereotype, which resulted into the denial of the existence of gangs in Europe (Gatti, Haymoz, Schadee 2011;Klein et al. 2001: xii).
Already in the 1930s the gang is defined by Trasher, a significant American author; the definition reads: "The gang is an interstitial group originally formed spontaneously, and then integrated through conflict.It is characterized by the following types of behavior: meeting face to face, milling, movement through space as a unit, conflict, and planning.The result of this collective behavior is the development of tradition, unreflective internal structure, esprit de corps, solidarity, morale, group awareness, and attachment to a local territory" (Trasher 1927, cit. Bjerregaard 2002: 33).Gang definitions according to Miller (1975, cit. Ball, Curry 1995) or Curry andSpergel (1988, cit. Siegel, Welsh 2011) followed.Also some more extensive studies on gangs were drawn up further to Thrasher ' s work: for instance Delinquent Boys (1955) by Albert Cohen and Delinquency and Opportunity (1960) by Richard Cloward and Lloyd Ohlin.In the 1990s, it was Ball and Curry (1995) who dealt with the issue of defining the gang.According to them, definitions of a certain term are often worded with respect to their future function and specific use (qualitative research of life in a gang; international comparison of gang members ' prevalence rates, etc.).Hence, no wonder that it is an uneasy task to define the gang.While one group of researchers still seeks a universal gang definition allowing its research across continents as well as cultures, others point out that such a definition cannot be unequivocally worded.For example, Haymoz (2010) is convinced that preferring one or another definition to the others results in obtaining completely different conclusions within various delinquency studies.Researchers should thus seek to formulate a single general gang definition permitting a more detailed research of the phenomenon in question on an international level.However, Petersen (2000) warns that one single gang definition will never exist as the gang nature differs in each society, and, besides, each researcher has a different idea of how such groups should be defined.
In terms of content, we have to ensure that the gang definition is neither too broad nor too narrow (Robinson 1954;Ball, Curry 1995).Both extreme cases usually result in including other social groups or, to the contrary, in ignoring a certain part of reality.This problem has been often dealt with by means of creating various gang typologies (e.g.Cloward and Ohlin ' s typology) (Ball, Curry 1995).Similarly, Medina et al. (2013) prove that also the gang definition according to the Eurogang group (see below) and its subsequent operationalization may lead to the identification of a separate group, the typical activity of which is recreational drug use.The authors believe that such individuals can be hardly classed as a gang by criminological benchmarks, and it is therefore desirable that they are either fully excluded or analyzed separately from delinquent gang members.
Many studies focusing on gang definition have also argued about the suitableness of identifying gang members based on delinquent activities.As Ball and Curry (1995) point out, definitions including illegal activities tend to minimize any further theoretical differences among gangs other than just delinquency rates.Characteristics such as friendship or common interests, generally considered as positive, are thus left out or directly excluded from the gang definition.Also Bjerregaard (2002) advises against including illegal activities into gang definitions.In many respects, their inclusion presents a tautology as we examine gangs especially in order to explain crime or delinquency in society, so we cannot consider criminal behavior as one of the main characteristics of a gang.On the contrary, according to Klein and Crawford (1967) society does not condemn gangs because of their normal behavior.It is the "delinquent product" of such a group that arouses a social reaction.Hence, the group ' s delinquent nature may be considered one of the main criteria, indispensable for defining a gang member.
Currently, Malcolm Klein belongs among the most important authors dealing with gang membership.At the end of the 20th century, the Eurogang group was created under his leadership.Based on a number of common workshops, the group has defined the main characteristics of a gang: group nature, duration in time, spending free time in public places, existence of group identity and committing illegal activities.A gang is "any durable, street-oriented youth group whose identity includes involvement in illegal activity" (Weerman et al. 2009: 20).
This article is dedicated to one of the most current gang definitions, the formulation of which has been connected with the Eurogang group.The definition has been operationalized and subsequently integrated into ISRD-2 and ISRD-3, i.e. two waves of International Self-Report Delinquency Study organized in the Czech Republic.Discussions on the appropriateness of the definition started already upon implementation of the ISRD-2 research; at the same time there were first analyses seeking to expand the original battery of questions -offered by the Eurogang group to identify a gang member -by other possible indicators (Gatti 2010).On the contrary, other studies suggest that even a single question can measure a respondent ' s gang membership in a positive way (Esbensen et al. 2001).This issue is discussed in further detail in the first part of this paper, being followed by an evaluation of the delinquent nature of all previously defined gang groups.The final part hereof presents an analysis of factors allowing the best prediction of the gang membership.Differences among the analyzed groups could suggest how the given gang definitions work and what kind of respondents are identified based on them.

International Self-Report Delinquency Study (ISRD)
The ISRD aims at deepening current knowledge in the area of juvenile delinquency as well as at enabling comparative studies on the said issue.In 2006-2007, 33 countries, including the Czech Republic, took part in the second wave of the study.The study focused on primary and secondary grammar school pupils aged approximately 12 to 16 years (seventh, eighth and ninth grades).A questionnaire form containing questions about family, school, neighborhood, leisure time spending, delinquent friends, attitude toward violence, self-control, victimization and the delinquency as such has been created in order to maximize comparability of international data (Podaná, Buriánek 2007).It was the first time that indicators, which were developed by the Eurogang group based on the above-stated definition, were included in the study to measure gang membership.Also, the third wave has been realized in a similar way.Although there have been a lot of modifications in the questionnaire, the indicators measuring gang membership remained unchanged.1Data for the Czech Republic were collected in spring 2013; however, this stage of the study is still pending in some participating countries.Hence, we have to do with the most current data describing the delinquency of Czech youth.
A national representative sample with an oversample for two cities, Prague and Pilsen, was opted for both the study waves.As ISRD is preferentially focused on data collection at city level (Steketee, Moll, Kapardis 2008), data on the two cities mentioned are used in the first part of this study, presenting prevalence rates of gang delinquency.Such data are suitable for direct comparison of prevalence rates with other countries participating in the study.The full unweighted data set are used for multinomial logistic regression where the increased number of respondents may significantly influence precision of the regression coefficient estimates.Table 1 shows respondent numbers for both ISRD waves.A total of 3,245 respondents took part in the 2006-2007 study, 725 respondents in Prague and 500 in Pilsen.In 2013 the sample amounted to 3,462 respondents, 855 of them were from Prague and 834 from Pilsen.Both samples included approximately a half of women and a half of men.The pupils ' average age oscillated at about 14 years.In most cases, respondents were represented by pupils born in the Czech Republic and attending primary school.

ISRD Gang Definition and Possibilities for its Revision
According to the Eurogang group, there are four core questions measuring gang membership.A respondent was identified as a gang member under Eurogang if he/she answered affirmatively all the following questions: 1.Does this group spend a lot of time together in public places like the park, the street, shopping areas, or the neighborhood?2. How long has this group existed? 3. Is doing illegal things (against the law) accepted by or okay for your group? 4. Do people in your group actually do illegal things (against the law) together?
Although the Eurogang group uses only the said indicators (core definitional questions) to identify a gang member, they have suggested also other, complementary questions (descriptors), the use of which is recommended for a more detailed gang research.One of them is also the so-called self-identification question: "Do you consider your group of friends to be a gang?"This question is not indispensable, according to Eurogang, for deciding on a respondent ' s gang membership, but it can be seen as a complement to the other four."It can be used to see whether gang members perceive themselves as such" (Weerman et al. 2009: 30).
However, Gatti ' s research (2010) indicates that this question could represent an important link in identifying a respondent as a gang member.While existing research has been oriented at differentiating respondents as gang members and non-members, Gatti assumes the existence of a continuum of "gangness", i.e. the existence of social groups showing some, but not all, gang characteristics.He measures the existence of that continuum using the Mokken scale; in addition to the four core Eurogang gang membership indicators, his analysis includes also the complementary self-identification question.The Mokken scale is a hierarchical scaling method similar to the Guttman scale, assuming the existence of a latent variable.It is represented by a series of indicators closely related to the latent variable.The indicators as such must have a hierarchical nature: we can expect that a respondent who answers a "more demanding" question affirmatively will also answer those questions that are "less demanding" affirmatively.A respondent ' s individual score on the scale is represented by his total number of affirmative answers (Van Schuur 2003).
Analysis results shown in Table 2 indicate the existence of a scale measuring one latent variable: gang membership.Loevinger ' s Hi3 coefficient shows the number of errors for individual questions; an error means an affirmative answer to a "more demanding" question and a negative answer to a "less demanding" question.If the Hi coefficient stands for a value lower than 0.30, the indicator should be eliminated as it does not form a continuum homogeneous with the other questions.Nevertheless, in our case, also the coefficient for the last question -"Do you consider your group of friends to be a gang?"equals 0.44, and the question has thus its significance in terms of identifying a respondent as a gang member.That is why Gatti (2010) recommends that the question should be included among the above-stated questions defined by the Eurogang group.
Some authors further state that the self-identification question itself could very well define gang members.This form of definition is used in the United States.The Eurogang researchers avoid it mainly for their fear that the expression "gang" might not be understood by all study participants, and they point out that the appropriateness of this technique may be questionable in other countries despite the fact that it has been quite successfully applied in the USA, where the majority of people understand very well what a gang means (Weerman et al. 2009).And indeed, Esbensen et al. (2001: 123), who car-ried out a study on students in 11 American states, concluded in their paper that "the simple question 'Have you ever been a gang member?' was understood by the respondents in such a manner that one can surmise that there exists a shared understanding of what this term means, not only by former or current gang members, but also by nongang youth".They consider the self-identification technique as an especially robust gang membership measure that allows dividing respondents in gang members and non-members (Esbensen et al. 2001).We decided to verify such assumptions by comparing the gang members defined only based on that question with those defined based on Eurogang criteria or the Mokken scale.
The understated chart compares the numbers of ISRD-2 and ISRD-3 gang members, whereby all the above-stated definition approaches have been taken into account.Gang members were identified based on Eurogang and based on the Mokken scale outcomes as well as based on the self-identification question.Each respondent who answered affirmatively all of the four above-stated questions was identified as a gang member under the Eurogang definition.A respondent who -in addition to that -gave also an affirmative response to the self-identification question was identified as a gang member in line with the Mokken scale.A respondent who gave the affirmative response but answered at least one of the four Eurogang questions negatively was assigned to the third group to be analyzed, the so-called self-identified gang members.Graph 1 shows that the highest number of gang members is identified in accordance with the Eurogang logic both within the ISRD-2 (8 percent) and the ISRD-3 (7 percent) data.The other two approaches result in lower numbers.While the number of gangs slightly declines using the Mokken scale, there has been an increase in those who termed their group a gang, without any account to their answers to the remaining questions.As boys predominate in the respondent group defined this way, we can assume that they find it less difficult to call their group a gang, or they may be even doing so because they find such a term very attractive.
Consequently, we asked why these respondents were not identified as gang members under the Mokken scale.The most frequent variants are shown in the Table 3, demonstrating that more than a half of the respondents (52 percent) declare that they do not accept or/and do illegal things (Eurogang questions no. 3 and no. 4, see above).Almost one-fifth (18 percent) of respondents state that in addition to not doing illegal things, they do not spend a lot of time together in public places.A lower incidence of delinquent activities (of self-identified gang members) is also further evidenced by the delinquency indicator.In terms of describing the above-mentioned groups, it might be interesting to see how such groups differ from those respondents who have a group of friends but do not meet the conditions to be assigned to one of the gangs in question.Hence, the following text offers a comparison of delinquency (overall and by types of acts) committed by the groups being analyzed, as well as a comparison of factors that may predict membership in such groups.The factors will be divided in individual (self-control, personal morality) and social (family, school, risk behavior).

Delinquent Nature of Gangs
Only delinquent acts that were measured by an identical or a very similar question in both study waves have been compared in order to avoid possible misinterpretations or distortions of results.A total of 11 delinquent acts were subject to comparison: intentional damage to property; theft in a shop or in a supermarket; burglary; bicycle theft; motorbike or car theft; car break-in; using a weapon or force to get hold of money or some property; carrying a weapon; participation in a group fight; hitting or hurting someone with a stick resulting in an injury; and drug dealing.Committing at least two delinquent acts in a recent period (12 months in our case) represents the main delinquency indicator.Such an a indicator focuses especially on individuals delinquent at present.The twoact limit guarantees that the act has not represented only a sporadic and thus a probably random incident (Podaná, Buriánek 2007).Graph 2 presents a comparison of delinquency indicators for all groups analyzed, including respondents from other groups both for the ISRD-2 and the ISRD-3 data.Data indicate that the Eurogang members ' delinquent nature has been even more significant than in 2006-2007.The same indicator has slightly decreased for Mokken scale gangs.Especially the self-identified gang members can be expected not to commit delinquent acts to such an extent and as frequently as members of other analyzed gang groups even though they identify themselves and their group of friends as a gang.On the whole, gang members defined under the Mokken scale appear to be the most delinquent ones (44 percent for ISRD-2 and 41 percent for ISRD-3) while there are considerable differences in delinquency rates of gang members defined in line with different approaches.
Prevalence of delinquent acts has been calculated in order to find out which of them can best characterize the respondent groups being analyzed.The calculation has been based only on ISRD-3 data, which reflect the current level of prevalence in a better way. Figure 3 shows that especially the following acts are frequently committed by gang members: shoplifting; vandalism; group fight; carrying a weapon and drug dealing.In relation to all the acts mentioned, gang members are always more delinquent than other groups of friends.Nevertheless, it is especially the gang members defined under the Eurogang criteria or under Mokken scale who are engaged in delinquent activities.The latter score high also in less frequent delinquent acts such as assaulting or hurting a person with a stick resulting in an injury (12 percent); using a weapon or force to get hold of money or a thing (7 percent); or motorcycle or car theft (5 percent).However, it needs to be pointed out that in the said cases the analysis is based on a very limited number of respondents.

Factors Predicting Gang Membership
Only ISRD third wave data were used to calculate factors providing the best prediction of membership in the gang types being analyzed as those data are more up to date and thus better reflect current developments in the field.
Dependent variable.In our case, the dependent variable is represented by the abovementioned respondent groups in question, i.e. gang members in accordance with the Eurogang group; self-identified gang members; gang members identified based on the Mokken scale; and members of other respondent groups.
Controlled variables.Respondents ' sex and age present controlled variables; sex has been recoded as a binary variable where the 1 value is assigned to men, and age feeds into the analysis as a cardinal variable with a minimum of 11 and a maximum of 17 years of age.
Individual factors.Individual factors include an individual ' s self-control and personal morality, i.e. prosocial behavior of an individual.The ISRD-3 measures self-control using a battery of nine questions.The scale is internally consistent, which is proven by Cronbach ' s alpha (0.79).It feeds into the analysis as a weighted summation index where 1 and 4 mean an individual ' s low and high self-control, respectively.Also personal morality, measured by a battery of eight questions (Cronbach ' s alpha = 0.75), feeds into the analysis as a weighted summation index with the values equalling to 1 and 4 for strong and weak personal morality, respectively.
Social factors.Family structure (1 = single parent family), family bonds, truancy (1 = yes), and an individual ' s risk behavior in his/her leisure time were included among social factors that might affect gang membership.All the variables mentioned except for the one measuring parent-child relationship quality are of binary nature.Parent-child relationship has been measured using a weighted summation index where 1 means that a respondent gets on very well with his/her parents (or a single parent) whereas 5 expresses very bad relationships within the family.An individual ' s risk behavior has been measured through questions about his or her habitual leisure time activities.Activities such as fighting with friends and seeking conflicts with other peers, doing some forbidden things, using alcohol or drugs, and hassling or frightening others just for fun have been taken into account.

Results
Table 4 presents the results of multinomial logistic regression.The reference category is represented by respondents from other groups, and all the results interpreted below shall be related to this respondent group.We believe that comparing gang members defined by us with this respondent group brings more advantages than their comparison with respondents who are not closer specified.Thus, the analysis was limited to the issue of social groups and their members.
The overall model is significant and explains almost one-third of the variability of the dependent variable (Nagelkerke R 2 = 0.285) provided that all the variables mentioned above are included.Model 1 operates only with controlled variables; Model 2 analyses the individual factors as well; and Model 3 represents the overall model predicting gang membership.
While in the case of the first two gang groups in question sex remains significant even if individual and social factors are taken into consideration, the said does not apply to gang members defined under the Mokken scale.Consequently, there are probably no differences in gender composition between this group and other groups of friends.The fact that women have a greater chance to become members of a gang defined in line with the Eurogang group (OR = 0.585) whereas it is mainly men who identify themselves as a gang (OR = 1.671) has been found interesting.One of plausible interpretations may be that it is rather men who consider calling their group a gang as a means to increase their social status among peers, and identifying themselves as such is therefore more attractive for them than for women, who are more careful and might have more fear of such a term.
The theme of self-control has been dealt with by Gottfredson a Hirshi (1990).According to them, an individual with low self-control is characterized especially by high impulsiveness and insensitivity.He/she is focused on short-term goals and on physical, non-verbal and risky behavior, resulting in his/her involvement in crime.Low self-control in combination with opportunity to commit (an often not planned) crime presents the underlying cause of delinquent behavior.Model 2 highlights the significance of individual characteristics in terms of gang membership.Low self-control and weak personal morality are typical for all gang groups.Individual factors also remain significant in the final model, the overall one (Model 3).There is only one exception: respondents who did identify themselves as gang members.In this case, self-control is no longer as significant as in other gang groups.Nevertheless, it still remains marginally significant: α < 0.1.Consequently, the statement that respondents who show low levels of self-control and weak personal morality have a greater chance to be involved in a gang than in other peer groups seems to be valid.
Significance of family bonds has been already pointed out by, among others, Aebi (2010) or Matoušek and Kroftová (2003).According to Aebi (2010), the significance of family structure extinguishes if the family dynamics (relationships between family members) is identical or at least similar in all family types (complete, monoparental, composed).In other words, what is important in relation to an individual ' s delinquent behavior is not whether his or her family consists of both or just a single parent, but whether the monoparental and composed families resemble complete families in terms of the quality of family bonds.Matoušek and Kroftová (2003) suggest that an individual ' s personal development significantly depends on the quality of bonds between the child and its parents and/or other family members.Our data confirm such assumptions to a great extent.While the quality of family bonds considerably influences whether an individual becomes a gang member, family structure remains insignificant for all the groups.The only exception is represented by self-identified gang members where the variable measuring family bonds appears to be insignificant as well.Leisure time risk behavior represents the main social factor predicting membership in such a group.A person manifesting risk behavior has a three times higher chance (OR = 2.959) to become a self-identified gang member.
Factors predicting Eurogang gang membership and Mokken scale gang membership are identical in the majority of cases.It is only truancy that presents an exception: while this variable is significant for the former respondent group, the same does not apply for the latter one.Those who have experienced skipping of school classes demonstrate an almost 1.5 times higher chance to become members of a gang defined in line with the Eurogang group (OR = 1.458).And it is truancy that poses one of current problems related to juvenile delinquency.Garry (1996), among others, states that pupils playing truant show a higher chance to get involved in illegal activities, such as drug abuse and trafficking, alcohol intake or committing violence.The overall results demonstrate an intersection of three core factors predicting gang membership: self-control, personal morality and risk behavior.

Discussion and Conclusion
This study looks into current trends in defining gang members.Three groups of gangs were defined based on previous research, and they were then compared with other respondent groups by means of the prevalence rates of delinquency and the factors that can predict the said membership.The paper aimed at explaining the difference between members of the gangs in question and respondents who do have a group of friends, but who were not identified as gang members.
The data analysis has shown that the Eurogang definition generates somewhat higher number of gang members than other two gang definitions.An interesting find is that the difference in the extent of delinquent activities committed by individual respondent groups appears to be statistically significant in both study waves.The most frequent delinquent activities include shoplifting, vandalism, group fight and carrying a weapon as well as drug dealing.The fact that respondents who identified their group as a gang do apparently not commit as many delinquent activities as other gang groups has been proven by both the prevalence rates calculated and the analysis of "non-affirmative responses", which shows reasons of elimination of such respondents from gangs defined using the Mokken scale.This supports our assumption that self-identification may play the role of an attractive label in terms of gang membership.Nevertheless, this respondent group is still involved in delinquent activities to a considerably higher extent than other groups of friends.The most delinquent respondents are found within Mokken scale gangs.
Some partial assumptions resulting from other studies on similar topics have been confirmed also for the youth in the Czech Republic by means of a multivariate analysis.Low self-control and weak personal morality are connected with a higher degree of delinquency.Family structure does not play an important role in the prediction of gang membership; it is the quality of family bonds that can be considered as a significant factor.Analysis indicates that poor family bonds predict gang membership in all gang groups except for the one in which the members did identify themselves as a gang.This group membership is influenced especially by the respondents ' risk behavior in their leisure time.This variable together with personal integrity and self-control appears to be significant for all the gang groups analyzed.
We have been aware of the two main limits of this paper: lower respondent numbers in individual gang groups feeding into the analysis, and the narrowing of results of delinquency prevalence rates only to urban gangs.In terms of research into gangs, respondent numbers present a ubiquitous problem.Lower rates may have a negative effect especially on the results of multinomial regression, more precisely on odds ratio (OR).That is why we decided to use the full unweighted set of ISRD third wave data.However, higher confidence intervals have been found in relation to the variable measuring respondents ' risk behavior.It is thus necessary to take into account that in this case the odds ratio may be distorted to some extent.Nevertheless, the outcomes presented seem to be very logical, and they correspond to conclusions of other studies dealing with gangs and meth-odological issues of their definition (for instance, Esbensen et. al. 2001;Haymoz 2010;Moravcová 2012;Matsuda, Esbensen, Carson 2012).
Furthermore, only respondents from two large cities in the Czech Republic, Prague and Pilsen, were included in the analysis of delinquency prevalence.This was done especially because the ISRD study is preferentially focused on data collection in cities (a minimum of two big cities in each country; what is meant by "big city" is left to researchers in the given country and depends on the relative importance of a given city or town).Although participating countries could opt for a national representative sample, which is the case of the Czech Republic, the majority of them have collected or will collect data at the city level.Hence, we have limited the analysis of delinquency prevalence only to such data in order to ensure a higher comparability of results presented herein with international data.Besides, gangs are often considered to be a predominantly urban phenomenon.
Defining gangs and their members poses an uneasy task, which often depends on the researcher and the issue being dealt with at a given moment.A situation favorable to the adoption of a single universal gang definition will be rare.Therefore, already at the very beginning of an analysis a thorough consideration should be given to the selection of a definition approach to be applied as each type of definition may generate a different group of respondents showing different characteristics and may thus lead to completely different results.

Graph 2 :
Gang members according to the delinquency indicator -at least two delinquent acts in a recent period (%) Note: N ISRD-2 = 1225, N ISRD-3 = 1689; two-sided 95% confidence intervals.Respondents who declared they had no group of friends are not included in the chart.Source: ISRD-2 and ISRD-3 Prevalence of delinquency for individual gang groups (%) in ISRD-3 Note: N ISRD-3 = 1689.Respondents who declared they had no group of friends are not presented in the chart.Source: ISRD-3

Table 1 :
Basic characteristics of samples

Table 2 :
Using the Mokken scale to evaluate the process of a respondent ' s identification as a gang member within ISRD-2 2 Source: Gatti ' s presentation at the EUROCRIM conference, 10 Sept 2010.Analysis of 63 large and medium-sized cities in 30 countries participating in ISRD-2, N = 40 678.

Table 3 :
Grounds for not assigning respondents to a gang under the Mokken scaleThe group does not spend a lot of time together in public places and, at the same time, does not accept and do illegal things