“In Police We Trust?”—Confidence in the Police Among Adolescents With and Without a Migration Background in Germany

Overall, adolescents exhibit a high level of confidence in the police. Nevertheless, it is problematic when certain social groups show lower levels of confidence in the police as this can affect police legitimacy. Based on a representative survey of ninth-grade students in Lower Saxony (N = 12,444), conducted in 2019, the current study examines confidence in the police among adolescents with and without a migration background. As well as an examination of the correlation, the study analyses whether the lower rate of confidence in the police among adolescents with a migration background can potentially be explained by them being exposed to a greater share and burden of influencing factors which impact confidence in the police (e.g. police contact, involvement with delinquent peers and neighbourhood characteristics). It is shown that adolescents with a migration background exhibit lower levels of confidence in the police than adolescents without a migration background. Although the correlation decreases upon the inclusion of other factors that influence confidence in the police, this relationship cannot be fully explained by the fact that young people with a migration background are also more burdened by the other determinants. However, the strongest correlations with confidence in the police are found for affinity for violence, friendship with delinquent individuals, police contact, property offenders, fear of crime and incivilities.


Introduction
Police trust or confidence in the police represents a central foundation for a democratic rule of law (Kääriäinen, 2007;Leitgöb-Guzy, 2018) because perceptions of just, fair and respectful treatment by the police have a direct impact on their legitimacy Oberwittler, 2016). According to procedural justice theory (Tyler, 2011), legitimacy is an important prerequisite for successful policing. For example, it is important for the police that victims and witnesses report crimes and make police-relevant statements (Leitgöb-Guzy, 2018). Furthermore, adolescents who confer legitimacy on the police are less willing to break the law .
The everyday life of the police is fundamentally characterised by the issuing of orders and commands, as well as interventions regarding the physical integrity of individuals. This involves both the exercise of laws and the use of force (Kipping, 2014: 98). Although the use of force is democratically legitimised, it can also be applied disproportionately, especially if it has a stigmatising effect (Lehmann, 2016). The Rotterdam Charter states the following about the main tasks of the police: The main objective of this Charter is to enhance organisational commitment to the issue of policing a multi-ethnic society. The first requirement is a clear mission statement for the organisation specifying its commitment to equal treatment and to combating racist and xenophobic violence. This should precede any tangible measures. (The Rotterdam Charter: Policing for a Multi-Ethnic Society, 1999: 366) In the last year in particular, the issue of police violence against the population, especially against BIPoC, as well as police-directed violence from the population, has come more to the foreground around the world as well as in Germany. The decisive factor in this was not only the killing of George Floyd in the USA, but also the debate regarding police violence and racism within the German police (e.g. the discovery of racist police What-sApp chats and the circumstances of Oury Jalloh's death in police custody, which remain unclarified). Such discussions led to the demand for independent studies on racist attitudes in the German police (for a comparative and historical analysis of the police in Germany in comparison to the USA, see Fairchild (1988)).
Based on these previous debates, it could be concluded that the confidence of German people in the police is extremely poor. However, studies show that the vast majority of adults in Germany trust the police (e.g. Birkel et al., 2014;Landeskriminalamt Niedersachsen, 2018;Riesner & Glaubitz, 2020); however, confidence in the police seems to be lower than in the USA (Cao, 2001). For example, in the survey by Riesner and Glaubitz (2020), more than 90% of respondents show a rather high or high degree of confidence that the police act according to the rule of law. Survey studies show that young people also express a high level of confidence in the police, with over 77% of students declaring high confidence in the police (Krieg et al., 2020). However, evidence from the USA shows that this rate is lower than it is for adults . The fact that adolescents are constrained by more rules, which therefore creates more opportunities for negative interactions with the police, may provide an explanation as to why adolescents have less positive attitudes towards the police than adults (Payne & Gainey, 2007). According to , targeting adolescents' police trust is important because, firstly, they represent a significant proportion of the population that is contacted and arrested by the police (for Germany, see Bundeskriminalamt, 2020; for the USA, Snyder & Sickmund, 2006) and, secondly, police are usually the first and only criminal justice personnel with whom adolescents have contact.
Despite the high level of youths' confidence in the police, it is nonetheless problematic when certain social groups show lower levels of confidence in the police. Ethnicity is one of the most common demographic characteristics associated with perceptions of the police (Schafer et al., 2003). Previous studies suggest that differential perceptions of the police exist in part because of differences in the way people of certain ethnicities are treated, or believe to be treated, by law enforcement officers (Payne & Gainey, 2007;evidence from Germany: Amnesty International, 1995). It is important to remember that individual experiences of police contact are generally symbolically perceived as a collective encounter between the group of origin and the state. If police officers treat certain ethnic groups unfairly or disrespectfully, this can be interpreted as implicit disrespect for the entire social group (Oberwittler, 2016;Tyler & Blader, 2003).
The current study aims to investigate whether confidence in the police differs between adolescents with and without a migration background (for a critique of the term, see El-Mafaalani, 2017). In addition, it will be analysed whether this correlation can be explained by the fact that adolescents with a migration background are more affected by other factors influencing confidence in the police, which are discussed within the state of research. As for the differentiation between the terms "trust" and "confidence", we decided to use the word "confidence" since it is more likely to be associated with an evaluation of the police as an institution rather than individual police officers (Cao, 2015).

Ethnicity and Migration Background
Since the late 1960s, studies have linked ethnicity to evaluations of the police. An overarching finding of these studies is that while Black people do have positive perceptions of the police, these are lower than the evaluations of White people (Ackerman et al., 2001;Dowler & Sparks, 2008;Jacob, 1971;Scaglion & Condon, 1980;Smith & Hawkins, 1973;Smith et al., 1999). This finding is also evident, albeit weaker, among adolescents (Brick et al., 2009;Hurst et al., 2005;Leiber et al., 1998;Schuck, 2013;Wu et al., 2015) and when controlling for neighbourhood characteristics and contact experiences with the police (Reisig & Parks, 2000;Schafer et al., 2003;Schuck et al., 2008). Regarding the link between migration background and perceptions of the police, such a clear picture cannot be drawn. There are some studies that cannot identify any association between migration background and evaluations of the police (Birkel et al., 2014;Leitgöb-Guzy, 2018;Riesner & Glaubitz, 2020). For adolescents,  indeed show that migrants have more negative perceptions of the police, but this association disappears when controlling for economic and social integration. In their further analysis, they find that it is not the immigrant background per se, but primarily the status as a visible ethnic minority that is crucial. Oberwittler (2016) also finds no differences in general attitudes towards the police between young people with and without a migration background in Germany. Regarding differential attitudes, however, adolescents with a migration background in Germany are more likely to agree with the two statements "The police are disrespectful towards young people" and "The police treat foreigners worse than Germans". Roux (2018) states that unfair treatment by the police is perceived to a stronger degree by ethnic minority than ethnic majority pupils, even when controlling for neighbourhood effects, offence-related police contact and attitudes towards discrimination. Similarly, other studies conclude that people with a migration background evaluate the police more negatively (Baier et al., 2019;Baier & Ellrich, 2014;Landeskriminalamt Niedersachsen, 2018). Overall, no clear result can be discerned regarding the correlation between migration background and confidence in the police.

Gender and Education
In relation to gender, previous research shows different results regarding confidence in the police. While some studies state that women have a more positive perception of the police (Cao et al., 1996;Landeskriminalamt Niedersachsen, 2018;Leitgöb-Guzy, 2018;Taylor et al., 2001;Weitzer & Tuch, 2002), there are also studies that point to a more positive police assessment among men (Birkel et al., 2014;, or at least find no correlation when controlling for other factors (Brick et al., 2009;Payne & Gainey, 2007;Riesner & Glaubitz, 2020;Wu et al., 2015).
Some studies find greater levels of confidence in the police among more highly educated individuals (Leitgöb-Guzy, 2018;Reisig & Parks, 2000) or less educated individuals (Cao, 2011;Groll, 2002). In other studies, the education effect is inconsistent (Schafer et al., 2003;Schuck et al., 2008) or no relationship is found (Cao et al., 1996;Groll, 2002;Payne & Gainey, 2007). Leiber et al. (1998) propose that it is important to look at lifestyle and involvement in delinquent subcultures in terms of attitudes towards the police. They contend that negative perceptions of the police form part of delinquent subcultures and that the greater the adolescent's involvement in the deviant subculture, the greater the likelihood of negative attitudes towards the police. In their analysis, they conclude that adolescents who engage in more delinquent behaviours and share deviant attitudes are less likely to respect the police (see also Brick et al., 2009;Chow, 2011;Lake, 2013;Schuck, 2013;Wu et al., 2015).

Delinquent Behaviour, Victimisation and Delinquent Subcultures
In accordance with the theory of differential association (Sutherland, 1947), involvement in delinquent subcultures by means of examining delinquent peer groups should be included in the analyses in addition to the previously mentioned factors (Brick et al., 2009). The theory thus states that delinquent subcultures are the result of interactions with delinquent individuals, whereby deviant attitudes and behaviours are transferred to others through exposure to deviant belief systems or operant conditioning (Brick et al., 2009). Although most subcultural theories do not explicitly discuss youth attitudes towards the police, they characterise the deviant subculture as one that includes norms that are negative or in direct opposition to respected adult society. From a subcultural perspective, young people hold negative attitudes towards the police not because of negative experiences with the police, but because they belong to a peer group that rejects all conventional forms of authority (Schuck, 2013). Empirical evidence confirms that adolescents who are more likely to spend time with delinquent peers exhibit more negative attitudes towards the police (Brick et al., 2009; see also Baz & Fernández-Molina, 2018). Furthermore, the consideration of victimisation experiences can be important in this context. For example, research shows that victims of a crime rate the police significantly worse than those respondents who have not yet had a victimisation experience (Birkel et al., 2014;Cao, 2011;Chow, 2011;Landeskriminalamt Niedersachsen, 2018;Payne & Gainey, 2007;Riesner & Glaubitz, 2020;Wu et al., 2009).

Contact Experience with the Police
If people's expectations of the police are not met, police contact may lead to less positive attitudes towards the police (Reisig & Parks, 2000). Differential perceptions of the police could thus also be a function of individual's experiences with the police and less a function of their demographic characteristics. The reason why young people or people with a migration background have a less positive perception of the police could therefore be that these groups have more contact with the police (Langan et al., 2001). Their more negative attitudes towards the police could be a product of the perception that they are the focus of excessive police attention (Schafer et al., 2003). Some studies conclude that mere contact experiences with the police have no effect on trust in police effectiveness (Birkel et al., 2014) or confidence in the police (Groll, 2002), while other studies find an association with more negative attitudes towards the police (Baz & Fernández-Molina, 2018;Brick et al., 2009;Cao, 2011).
In many studies, a difference only emerges when the type of contact, as well as the satisfaction with it, is considered. Voluntary contact (i.e. citizen-initiated contact) tends to have a positive effect, whereas police-initiated contact tends to have a negative effect (Brick et al., 2009;Schuck, 2013;Skogan, 2005). According to Schafer et al. (2003), however, the initiation of the contact does not play a role and instead, the only decisive factor is whether the contact was positive or negative. Studies are almost unanimous in demonstrating that negative contact experiences are related to more negative perceptions of the police (Brown & Benedict, 2002;Chow, 2011;Dean, 1980;Hurst et al., 2005;Leitgöb-Guzy, 2018;Reisig & Parks, 2000;Schafer et al., 2003;Schuck et al., 2008) with the exception of Staubli's (2016) study.
Regarding positive police contact experiences, varying results are reported. In some studies, positive contact experiences often cannot be associated with stronger confidence in the police (Leitgöb-Guzy, 2018;Schafer et al., 2003), trust in police effectiveness (Birkel et al., 2014), or perceptions of police service and fear of the police (Schuck et al., 2008). However, other research demonstrates associations between satisfactory police contact and greater trust in procedural justice (Birkel et al., 2014), positive general perceptions of the police in a neighbourhood and global context (Schuck et al., 2008), police trust (Staubli, 2016), satisfaction with the police (Reisig & Parks, 2000) and positive attitudes towards the police (Brown & Benedict, 2002;Hurst et al., 2005). Such differences in research results regarding positive contact experiences compared to negative contact experiences could be explained by a negativity bias (Schuck, 2013). Individuals may hereby assign greater weight to negative than positive experiences when forming attitudes towards the police (Skogan, 2006).
With regard to adolescents,  point out that adolescents who were stopped or arrested by the police and believed to have been treated badly held less positive attitudes towards the police, while adolescents who initiated contact with the police and perceived police behaviour as good were more likely to hold positive attitudes towards the police. However, police-initiated positive contact and adolescent-initiated negative contact are not related. Oberwittler et al. (2014) also illustrate that young people's opinions of the police only deteriorate with more frequent police contact.

Neighbourhood
With regard to the neighbourhood characteristics, two theoretical approaches can explain the influence on confidence in the police. On the one hand, disadvantaged neighbourhoods generally have a low level of confidence in the police because the residents blame the police for the poor housing conditions (Reisig & Parks, 2000). Particularly, those who have no direct contact with the police are more likely to base their assessments of the police on their neighbourhood culture, including perceived or actual crime rates (Jacob, 1971;Schafer et al., 2003). In addition, residents of disadvantaged neighbourhoods regard the police as an extension of the government and project their negative attitudes towards the government onto the police (Leitgöb-Guzy, 2018).

The Present Study
Since it is not possible to draw a clear picture of the link between adolescents' migration background and their perceptions of the police, the current study firstly employs analyses to examine whether adolescents with a migration background have less confidence in the police than adolescents without a migration background.
Hypothesis 1: Adolescents with a migration background have less confidence in the police than adolescents without a migration background.
As a next step, we analyse whether adolescents with a migration background are more affected by the common determinants of confidence in the police in terms of their involvement with delinquent peers, victimisation and delinquency, affinity for violence, police contact and neighbourhood characteristics.
Hypothesis 2: Adolescents with a migration background are more likely to be involved with delinquent peers, to be victims and offenders of a crime and to have a higher affinity for violence (H2a), have more police contact (H2b) and live in worse neighbourhood conditions (H2c).
As a next step, the study analyses whether a potentially lower level of confidence in the police among adolescents with a migration background can be explained by their hypothetically greater burden regarding other influencing factors of confidence in the police.
Hypothesis 3: Adolescents with a migration background have no longer less confidence in the police when the other determinants of confidence in the police are controlled for.

Analytical Strategy
Hypotheses 1 and 2 are tested using independent t-tests for metric or quasi-metric variables (two-tailed; prespecified significance level: p < .05) and chi-squared tests (significance level: p < .05) for nominally scaled variables. For hypothesis 3, adolescents' migration background is analysed within a linear regression model on confidence in the police (Model 1). Police trust is not normally distributed, as assessed by the skewness and kurtosis test for normality (p < .05). However, we did not choose to dichotomise the variable because dichotomisation always involves a loss of information. In addition, this assumption is mostly not necessary if the sample is sufficiently large, which is the case with our data. Clustered standard errors are necessary because the respondents were surveyed within their class groups. It can therefore not be assumed that the students responded independently, and this must be considered when calculating standard errors. As a second step, the analysis controls for sociodemographic factors such as age, gender and education (Model 2). Furthermore, involvement with delinquent peers, delinquent behaviour, victimisation and affinity for violence (Model 3), experiences with the police (Model 4) and neighbourhood characteristics (Model 5) are gradually included in five consecutive models.

Sample Extraction
The following analyses are based on the data of ninth-grade students in Lower Saxony, which were collected in 2019 as part of a periodically conducted representative criminological dark field survey by the Criminological Research Institute of Lower Saxony (Krieg et al., 2020). For the survey, school classes were randomly drawn (stratified sampling by school type) from all classes taught in the relevant school year (special educational schools with a focus other than learning were excluded). In 2019, 12,444 adolescents from 762 classes were surveyed, which corresponds to a response rate of 41.4% (50.8% male, mean age 15.1 years). Most of the adolescents in the sample of 2019 (68.9%) do not have a migration background, which means that 31.1% of the adolescents do have a migration background. This proportion is statistically significantly higher than in 2017 (27.7%; χ 2 (1) = 28.29, p < .001, φ = 0.04). The most common migration background for the ninth graders in the sample is Russian (6.1%). In 4.5% of the cases in 2019, the adolescents or their parents come from Turkey, and 3.9% of the cases from Poland. In addition, 1.5% of the adolescents have a Kazakh migration background, and 8.3% of the adolescents have an otherwise European migration background; most have parents who come from Southern Europe or come from there themselves (4.1%). Adolescents with an Asian migration background are represented by 4.7%, with most of them having a West Asian background (2.7%). The remaining ninth graders have a North American (0.4%), South American (0.5%), African (1.2%) or Australian (0.0%) migration background.
In the regression models, the sample is reduced due to missing values in the dependent and/or independent variables (if any variable contained at least one missing value, the case was excluded). Since the variables on fear of crime were only presented to a randomised quarter of the respondents, the sample is reduced to n = 2493 cases. The study was approved by the Lower Saxony State Education Authority. The adolescents' parents were informed about the study in advance and were asked to give their written consent for their child's participation. Prior to the survey, the students were informed that participation was voluntary, the answers would only be processed anonymously, the survey could be stopped at any time, individual questions could remain unanswered, and they would not suffer any disadvantages from non-participation. The survey, which lasted approximately 90 min, was supervised by trained test leaders and conducted in writing or on computers within the class group in the presence of a teacher.

Dependent Variables
Confidence in the Police Confidence in the police was measured using four different statements (see Table 1). The second statement is an inverse item, where high agreement corresponds to low levels of confidence in the police. To facilitate similar interpretation across all statements, the item was reversed prior to formation of the scale. Adolescents were able to rate their opinion on these statements on a scale from "1, disagree" to "4, completely agree". A mean value scale was formed using all four items, the reliability of which can be classified as acceptable with McDonald's Ω = 0.76.

Independent Variables
Migration Background A young person is considered to have a migration background (1) if they have a nationality other than German and/or were born in a country other than Germany, or if this applies to at least one of their parents. The respondent is only considered to have no migration background (0) in cases where "German" is ticked everywhere in the survey.

Gender and School Type
Gender is coded as "male (0)" and "female (1)". The school type is divided into "low (0)", "medium (1)" and "high (2)". The types of school can be classified according to the ISCED (International Standard Classification of Education). According to this categorisation, students of the low and medium school type reach lower secondary education (Level 2). Low school-type pupils usually complete 9 years of schooling, whereas medium school-type pupils usually complete 10 years of schooling. Pupils of the higher school type reach upper secondary education (Level 3) after 13 years of school.
Delinquent Behaviour, Victimisation, Affinity for Violence and Involvement with Delinquent Peers Experience of a violent crime (i.e. victim and perpetrator experiences) was measured with two variables by means of querying the following seven offences: (1) robbery, (2) extortion, (3) sexual harassment, (4) sexual violence, (5) bodily harm with a weapon, (6) bodily harm by several persons and (7) bodily harm by individual persons. If adolescents had committed at least one of these offences, they are considered to be a violent perpetrator (1) (no perpetrator (0)). Likewise, if the students had experienced at least one of these offences, they are categorised as a victim of a violent offence (1) (no victim (0)).
Measurements regarding victimisation and perpetration of property crime required queries into the four offences of (1) bicycle theft, (2) vehicle theft, (3) theft and (4) damage to property. Adolescents who had committed at least one of these offences are considered a property offender (1) (no property offender (0)). Likewise, students who had experienced at least one of the four offences are regarded as victims (1) (no victim (0)).
The adolescents' affinity for violence was assessed with the help of these four items: (1) The strongest must prevail; otherwise, there is no progress, (2) I would be willing to use violence to show what I am made of, (3) A bit of violence is simply part of having fun, (4) Without violence, everything would be much more boring. They were summarised into a mean value scale. Adolescents rated their answers on a scale from "1, untrue" to "4, very true". The reliability of the scale is rated as acceptable with a McDonald's Ω = 0.75.
In addition, the adolescents' involvement with delinquent peers was surveyed by asking how many of their friends had behaved delinquently in the last 12 months: (1) shoplifting, (2) robbery, (3) bodily harm, (4) property damage, (5) drug sales. Adolescents rated their answers on a scale of "1, zero delinquent friends", "2, one delinquent friend", "3, two delinquent friends", "4, three to five delinquent friends", "5, six to ten delinquent friends" and "6, over ten delinquent friends". A maximum value index was initially formed using the five items. If, for example, the respondents stated that none of their friends had committed shoplifting in the last 12 months, but two peers had sold drugs to others, the number of drug-dealing peers would be included in the overall index. This approach was taken because, in this study, it is not important in which way the adolescents' peers are delinquent as it is assumed that all types of delinquency are related to less confidence in the police. The reliability of the scale is rated as acceptable with a McDonald's Ω = 0.78. The overall index was then categorised into the response categories of "zero delinquent friends (0)", "one to five delinquent friends (1)" and "more than five delinquent friends (2)".

Contact with the Police
In order to measure contact with the police on the basis of illegal activity, adolescents were asked the following: "Have you ever had contact with the police because you did something illegal?" The young people answered "no (0)" or "yes (1)".

Neighbourhood Characteristics
The feeling of community safety (see Wilmers et al., 2002) was assessed with two questions on how safe or unsafe adolescents feel when they are outside in their neighbourhood/district (1) during the day and (2) in the evening, or at night. An average index was calculated from the data for daytime and evening/night. Adolescents rated their answers on a scale from "1, very unsafe" to "4, very safe". The two items correlate highly with each other (r = 0.57).
Measurements relating to the fear of crime were based on Bilsky et al. (1993). Adolescents were asked to indicate how often they feared (1) theft, (2) sexual violence, (3) robbery, (4) a terrorist attack and (5) assault in the last 12 months. A mean value scale was formed using these five items. Adolescents rated their answers on a five-point scale ("1, never", "2, rarely", "3, sometimes", "4, often" and "5, very often"). The reliability of the scale is rated as good with a McDonald's Ω = 0.81.
For the measurement of incivilities, adolescents were asked to indicate to what extent they agreed with the following statements: (1) there is a lot of rubbish lying around or it is dirty in the community, (2) there is a great need for renovation, (3) there are a lot of people drinking alcohol on the street, (4) there are a lot of people using or dealing drugs on the street and (5) people can generally feel safe. The last item was reversed for analysis. A mean scale was formed based on these five items. Adolescents graded their answers on a scale from "1, untrue" to "4, very true". The internal consistency of the scale is questionable with a McDonald's of Ω = 0.64, which means that the five items are not five homogeneous statements. In order to reflect the full range of content of the incivilities dimension and thus keep content validity high, a conscious decision was made to include different aspects of this dimension in the scale construction. It is necessary to use a sufficient number of different items to describe all relevant characteristics of incivilities.

Results
A high level of confidence in the police can be identified among adolescents. We see that over 80% of the adolescents agree or strongly agree with the statements "The police keeps us safe" (82.1%) and "The police try to help people" (80.2%) while also over 80% disagree or disagree strongly with the statement "The police treat people unfairly" (81.3%). Beyond that, 68.1% of the adolescents agree or strongly agree that they have "great trust in the police". Overall, 66.0% of the students can be classified as confident in the police because they reach an average value above > 3 on the mean value scale.
Nevertheless, the analysis shows that adolescents with a migration background (M = 2.93; SD = 0.65) have lower levels of confidence in the police than young people without a migration background (M = 3.12; SD = 0.61), which confirms hypothesis 1 (see Table 1). According to Cohen's (1988) convention, the effect size can be interpreted as small (t (12,017) = 15.68; p < .001; d = 0.31), but even small effects can be highly relevant, especially at the population level. Based on differentiations according to the individual statements, the correlation is particularly evident regarding the statement "The police treat people unfairly", which young people with a migration background agree to a significantly higher degree than adolescents without a migration background.
Further analyses show that all variables related to confidence in the police differ according to migration background, with the exception of fear of crime (see Table 2), which confirms hypotheses 2a and 2b and partly confirms hypothesis 2c. Adolescents with a migration background are thus slightly more likely to report having been perpetrators of property and violent offences, victims of violent and property offences, and to have had contact with the police as a result of illegal activity. In addition, adolescents with a migration background are more likely to report having delinquent friends, are more prone to violence and are more likely to notice incivilities in their neighbourhood. If the effect sizes are considered according to Cohen's (1988) convention, however, a practically significant effect is only observable for violent offenders, affinity for violence and the extent of incivilities. As a next step, it is analysed whether adolescents with a migration background being faced with the greater burden of other determinants that impact confidence in the police can explain the relationship between migration background and confidence in the police. We provided a correlation matrix for all included variables of the linear regression model in Table 4 in the Appendix section. Almost all of the correlations with confidence in the police were significant and of small to moderate size (r = − 0.34 to 0.10).
For linear regression analysis, migration background is included as an independent variable in a regression model (see Table 3). The analysis confirms that adolescents with a migration history have less confidence in the police than adolescents without a migration history (b = − 0.202; p < .001). In a second step, the other demographic factors are included in the analysis. According to this, girls have higher confidence in the police than boys. In addition, students from medium school types exhibit lower confidence in the police than students from high school types. The correlation of confidence in the police with migration background hardly changes. The explanatory power of the model slightly increases to r 2 = 0.035.
The third model includes the variables related to delinquency. The model shows that being a perpetrator of a violent or property crime and being a victim of a violent crime are both associated with significantly lower confidence in the police. In addition, adolescents with an affinity for violence and adolescents with delinquent friends exhibit lower levels  .212 .226 of confidence in the police. The significant correlation with the migration background remains but becomes noticeably smaller (b = − 0.105; p < .001). The other demographic characteristics no longer show significance due to the addition of these variables. Model 3 clearly gains overall explanatory power (r 2 = 0.201).
In the fourth step, police contact is added to the model. The analysis reveals that adolescents who have had previous contact with the police, as a result of illegal activity, have less confidence in the police. All other correlations remain largely constant. The explanatory power of the model slightly increases (r 2 = 0.212). The fifth and final model includes the neighbourhood variables. Here, a high level of reported incivilities and fear of crime are associated with lower confidence in the police. In the overall model, all included factors except for gender, school type, being a victim of a property crime and community safety are related to confidence in the police. A comparison of the standardised beta coefficients shows that affinity for violence, friendship with delinquent individuals, contact with the police, perpetration of a property crime, fear of crime and incivilities have the strongest associations with confidence in the police. Only then the migration background follows these associations in relation to correlation strength and ranks as the 7th (out of 8) strongest predictor.
Even when controlling for the other determinants, migration background is significantly related to confidence in the police (b = 0.092; p < .001), which contradicts hypothesis 3. However, we only find a weak association when the other variables are included. Lower levels of confidence in the police among adolescents with a migration background can thus be partially explained by different manifestations of the included factors within the two groups. But the fact that adolescents with a migration background are more affected by the other influencing factors cannot provide a complete account for the association between migration background and confidence in the police. Nevertheless, migration background still plays a role in the evaluation of confidence in the police, albeit a subordinate one. The explanatory power of the final model is r² = 0.226.

Discussion
Due to an ambiguous state of research and limited knowledge regarding adolescents, the present study investigated the relationship between migration background and confidence in the police among adolescents, while simultaneously controlling for other influencing factors of confidence in the police using a large representative student sample. The existing studies on this topic are mostly investigating adults and come to different conclusions. While some studies do not find an association between migration background and evaluations of the police (Birkel et al., 2014;Leitgöb-Guzy, 2018;Riesner & Glaubitz, 2020), others do find a significant association (Baier et al., 2019;Baier & Ellrich, 2014;Landeskriminalamt Niedersachsen, 2018;Oberwittler, 2016;Roux, 2018). The findings of the current study show that adolescents with a migration background have a lower level of confidence in the police than adolescents without a migration background. Although the strength of this relationship decreases to a weak association upon the inclusion of other factors that influence confidence in the police (especially involvement with delinquent peers), the migration background nonetheless plays a role in the assessment of confidence in the police, albeit in comparison with the other included determinants a subordinate one. The fact that adolescents with a migration background are more burdened by the other determinants of confidence in the police cannot, however, provide a comprehensive explanation for this result. Regarding adolescents with a migration background, their higher levels of delinquency, greater police contact and increased likelihood of growing up in neighbourhoods characterised by incivilities can partially account for their lower confidence in the police. These other factors that influence the confidence in the police are in some cases directly connected to police work (e.g. direct interactions in police contacts) and, in other cases, indirectly connected when considering police responsibilities such as fear of crime or incivilities. Further influencing factors are considered outside of the control of the police such as criminal offending. Altogether, it can be assumed that confidence in the police would improve if the principles of policing would be followed in every aspect.
Furthermore, it is also worth explaining the heavier burden of adolescents with a migration background regarding the other factors influencing confidence in the police. In addition to the normative developmental tasks of adolescents, adolescents with a migration background must cope with acculturation-related tasks that result from the encounter of two cultures. These tasks relate to socio-cultural and psychological adjustments and can include dealing with cultural differences, learning a new language (if necessary) and acquiring culturally appropriate values, beliefs, skills and behaviours, as well as developing ethnic and national identities (Oppedal, 2006;Phinney et al., 2001;Ward et al., 2001). In addition, individuals with a migration background often face acculturation-related stressors, such as discrimination and language problems (Titzmann et al., 2011). Due to these additional developmental tasks of adolescents with a migration background, higher levels of stress relating to a range of experiences can be expected.
Furthermore, the results suggest that the reasons for lower levels of confidence in the police among adolescents with a migration background should continue to be discussed. The small remaining effect can be due to factors that we did not include in our analysis such as a positive or negative experience with the police or perceptions of the media about police misconduct against individuals with a migration background. One possible explanation could relate to the differential treatment of adolescents with a migration background at the hands of the police, which, for example, was demonstrated by Oberwittler (2016) and Oberwittler et al. (2014) in relation to police identity checks, which are still understudied in Germany. These checks are linked to suspicions related to the location or situation but are not necessarily directed at an individual person in Germany (Oberwittler, 2016). Young people are often the target of identity verification checks that lack suspicious ground (Oberwittler, 2016). Such identity checks, which are independent of suspicion, offer the police great discretionary powers over whom they check (Oberwittler, 2016). In particular, the police's youth protection units explicitly approach young people. The practice of conducting identity checks that lack suspicious ground can lead to discrimination against certain ethnic groups when racial profiling is applied. This violates not only the German "Grundgesetz" (Art. 3 para. 3 GG) and the General Equal Treatment Act (Allgemeines Gleichbehandlungsgesetz; AGG), but also the prohibition of racial discrimination laid down in the European Convention on Human Rights and the International Convention against Racism (Thompson, 2020). Although Oberwittler (2016) and Oberwittler et al. (2014) did not find any differences between adolescents with and without a migration background in relation to the frequency of suspicionless identity checks carried out on young people, the behaviour of the police during these checks nevertheless differed. In cases of identity checks carried out on adolescents with a migration background, the adolescents' bags were checked significantly more often, and the police became somewhat more violent (Oberwittler, 2016). Likewise, adolescents with a migration background stated somewhat more frequently that the police provoked them or used violence during the period of contact. In terms of behavioural intentions, it was found that adolescents with a migration background were slightly less likely to report significant problems to the police . Further studies conclude that about 34% of young people with a Turkish migration background often feel very unfairly treated by the police (Heitmeyer et al., 1997; see also Salentin, 2008;Sauer, 2009). It is of high importance to raise awareness among the police about discriminatory effects relating to everyday activities and to counteract these (Lehmann, 2017). However, it must be simultaneously emphasised that attitudes and actions are a result of mutual perceptions and interaction processes (Gesemann, 2001). The focus in this study is based on information provided by adolescents. It therefore seems even more important to examine police attitudes within future studies to understand dynamics and processes, as well as to analyse mutual perceptions and derive actions from the results. This could provide a building block for tackling possible racial profiling in a preventive manner but could also contribute to strengthening the confidence of adolescents with a migration background in the police.
Furthermore, high-profile incidents of police misconduct can also negatively influence attitudes towards the police. Black people feel more affected, and the effect lasts longer, than for white people (Tuch & Weitzer, 1997). In a later study, the authors found that the effect of media coverage relating to police misconduct on police satisfaction was limited to Black people (Weitzer & Tuch, 2005). In a study by , perceptions of police misconduct (i.e. involving another person) constituted the strongest predictor of negative attitudes held towards the police.

Limitations
Overall, the present study has some limitations. For example, a very broad definition of migration background is applied, in that adolescents with a second-generation migration background are also included. Racialised individuals, in particular, experience discrimination in Germany and are more affected by incidents of police misconduct that attract media attention (Tuch & Weitzer, 1997;Weitzer & Tuch, 2005), which can in turn impact confidence in the police. A more differentiated definition would be beneficial, as racialization and visible ethnic minority status were not considered in this study (see . In addition, some relevant factors influencing confidence in the police are not included in the data set. Among other things, an indicator for neighbourhood cohesion is missing (see Cao et al., 1996). Moreover, only one measure of police contact could be included, provided that adolescents had done something illegal. Citizen-initiated contact (e.g. prevention work initiated by schools) was not surveyed. Furthermore, no distinction was made as to whether adolescents had a positive or negative contact experience.
Furthermore, it must be mentioned that the present study design only allowed the surveying of ninth-grade students, meaning that reliable statements are limited to this age group. As in all survey studies, the data may have been falsified by the respondents due to processes of social desirability or deliberate deception. Such processes, however, should have been kept to a minimum by providing the participants with detailed information about the underlying data protection mechanisms (e.g. anonymised survey, no feedback to parents, school or teachers). The response rate can be classified as satisfactory; however, it cannot be ruled out that particularly stressed pupils often did not participate in the survey due to health problems or truancy.
Finally, it must be pointed out that the present study is a cross-sectional study and can therefore only make statements about correlations, and not about causal relationships. Future studies should further investigate the relationship between, for example, involvement with delinquent peers and confidence in the police within the framework of longitudinal research.

Table 4
Correlation matrix of the independent and dependent variables (Pearson's r)  Table 4