Time matters: civic engagement and the development of anti-immigrant attitudes among adolescents

A recurring claim in the literature is that civic engagement in voluntary associations is crucial for the formation of a variety of interpersonal attitudes, such as social trust, tolerance, and intolerance. In this article, we study the role of membership in voluntary associations and volunteer work in the development of anti-immigrant attitudes among adolescents. Using a four-wave longitudinal design with a target sample comprising approximately 2000 adolescents between the ages of 13 and 16 at the first measurement occasion, this study contributes to previous research by analysing the impact on anti-immigrant attitudes of different dimensions of membership in voluntary associations and volunteer work. The results showed that membership in voluntary associations and volunteer work over several years serve to decrease anti-immigrant attitudes among adolescents over time. In addition, volunteering over several years seems to have a somewhat stronger impact on anti-immigrant attitudes than do various aspects of membership in associations. Our study also shows that anti-immigrant attitudes become more firmly established with age and that membership in only certain types of voluntary associations relates to attitudes toward immigrants. In sum, this study contributes to previous research focusing on the role of civic engagement in the development of interpersonal attitudes.


Introduction
A recurring claim in the literature is that civic engagement in voluntary associations is crucial for the formation of a variety of interpersonal attitudes (Putnam et al. 1994, pp. 88-89;Newton 1999;Putnam 2000, pp. 22, 288;Warren 2001). Through involvement in voluntary associations, people learn "the habits of the heart" (Bellah 1985) for example, social trust and tolerance), which are essential for the functioning of democratic society. In this regard, voluntary associations are assumed to have a socialization effect on the development of interpersonal attitudes (Newton 1999;Putnam 2000;Warren 2001; Morales and Geurts 2007;Cote and Erickson 2009). Indeed, some researchers also argue that the development of civic attitudes such as social trust and tolerance is an unintended "by-product" of civil society (Baggetta 2009). Confidence in the role of voluntary associations has resulted in policymakers emphasizing their importance in the strengthening of social cohesion in societies characterized by increasing ethnic and cultural differences (e.g., European Economic and Social Committee 2020).
However, empirical research in this area from the past two decades reveals a mixed and complex picture of the relationship between civic engagement and a variety of interpersonal attitudes, norms, and values. Some studies have, for example, revealed a positive relationship between membership in voluntary associations and social trust (e.g., Brehm and Rahm 1997;Paxton 2007;Stolle 1998;Wollebaek and Selle 2002;Jennings and Stoker 2004); tolerance (Rapp and Freitag 2015;Iglič 2010;Coffé and Geys 2007;Hooghe 2003;Cigler and Joslyn 2002;Almond and Verba 1963); and empathy (Bekkers 2005). Other studies, in contrast, have rejected the significance of voluntary associations and stress that the positive relationship between membership in an association and various forms of civic-oriented attitudes results from self-selection (Bekkers 2012;Quintelier 2013; Van der Meer 2016; Dahl and Abdelzadeh 2017). Furthermore, research shows that associations vary in how they relate to interpersonal attitudes (Stolle and Rochon 1998;Hooghe 2003;Hooghe andQuintelier 2013 2013). This suggests that the socialization effect of voluntary associations is group-specific and that involvement in certain groups may shape interpersonal attitudes while involvement in others does not.
At the same time, despite the endeavours of research, the extent to which membership in voluntary associations leads to the development of positive interpersonal attitudes remains unclear. There are several reasons for this. First, most previous studies have based their conclusions on cross-sectional data, which has inferential limitations in relation to dynamic processes. Second, longitudinal studies are few and those that are longitudinal often cover a relatively short period of time, up to two waves (Hooghe and Quintelier 2013). We know that many political and social orientations change slowly (Huddy and Sears 2011). Therefore, to determine the effect of involvement in voluntary associations on interpersonal attitudes, we need studies that cover a longer period. Third, previous studies focus mostly on formal involvement in associations, such as membership, and not on informal forms of civic engagement, such as volunteer work (see, however, Jennings and Stoker 2004). This is a weakness-especially given the rise of new and changing repertoires of civic and political engagement among younger generations (e.g., Sloam 2014). Fourth, to our knowledge, no study has focused on the long-term effects of membership in voluntary associations and volunteer work on attitudes of intolerance. All in all, given these limitations and the increasing xenophobia and negative attitudes toward immigrants in many countries, there is a need for a better understanding of the role of civic engagement in the strengthening of social cohesion in society. In sum, existing research does not allow for a definite conclusion on the effect of membership in voluntary associations and volunteer work on attitudes of intolerance.
This study aims to respond to the limitations of previous research by exploring the role of membership in voluntary associations and volunteer work on the development of anti-immigrant attitudes among adolescents. We use a unique four-wave longitudinal study from an extensive Swedish research project about young people's socialization. The target sample comprises approximately 2000 individuals between the age of 13 and 16 at the first measurement occasion. Participants lived in a medium-sized Swedish city with a total population of about 135,000, and as such several criteria relating to national representativeness were met (Amnå et al. 2009). By examining the longitudinal impact of different aspects of membership in voluntary associations and volunteer work on anti-immigrant attitudes, this study contributes to the overall discussion on the development of young people's political attitudes. Consequently, we aim to answer the following question: To what extent do membership in voluntary associations and volunteer work influence the development of anti-immigrant attitudes among adolescents?
The article is structured as follows: after this introductory section, we present our theoretical points of departure by drawing on the literature on the development of anti-immigrant attitudes among adolescents and the causal mechanisms of civic engagement and attitudes toward immigrants. This is followed by an explanation of the research design, including descriptions of our methods and research strategy. Thereafter, we present our results, describing development and changes in antiimmigrant attitudes over time. We also answer the question of how membership in voluntary associations and engagement in volunteer work impact the development of anti-immigrant attitudes among adolescents. Finally, the results are summarized and their implications for future research are discussed.

Theoretical setting and previous literature
Previous studies of political socialization show that early adulthood is a period when people become aware of the social and political world around them and when social and political attitudes develop and form (Krosnick and Alwin 1989;Sears and Brown 2013;Raabe and Beelmann 2011). Research shows that adolescence is a critical period in a person's life when attitudes such as social trust (Jennings and Stoker 2004;Stolle and Hooghe 2004;Abdelzadeh and Lundberg 2017) and tolerance  are open for change. This suggests that if we want to study the extent to which membership in voluntary associations and volunteer work influence the development of anti-immigrant attitudes, we need to analyse their effect during adolescence.
Research highlights how different contextual socialization factors may relate to the development of anti-immigrant attitudes among adolescents (Bennett and Sani 2008;Jones and Rutland 2018). Parents can affect their adolescent children through, for example, intergenerational attitude similarity and mechanisms of social learning (Bandura 1977;Jennings et al. 2009;Grusec 2011). A study by Gniewosz and Noack (2015) on German adolescents shows that parental attitudes can be used to predict their attitudes toward immigrants over time from early adolescence up to the age of 16. In addition, Miklikowska (2016) shows that the quality of the adolescent-parent relationship can affect the level of influence parents can have on their children. Adolescents who perceived their parents to be supportive showed higher parent-adolescent correspondence in terms of prejudice than adolescents who had low levels of parental support.
The views on immigrants held by their peers are another factor that can affect the development of anti-immigrant attitudes among adolescents. Titzmann et al. (2015) show that adolescents who established an intergroup friendship over the three time points demonstrated a significant decrease in negative prejudice. Similarly, van Zalk et al. (2013) found that friends' xenophobia predicted increases in adolescents' xenophobia. In addition, research shows that friendships with immigrants result in a higher degree of empathy, which in turn decreases intolerant attitudes toward immigrants (Miklikowska 2017). In many cases, the school serves as a central site for peer relations where adolescents from various ethnic and cultural groups interact, adapt to social norms, and find role models who influence the development of antiimmigrant attitudes (Benner et al. 2015;Özdemir et al. 2018;Miklikowska et al. 2019).
Voluntary associations are another important social environment where young people spend lots of time, and these can influence attitudes toward immigrants. According to the socialization perspective, voluntary associations and volunteer work enable people to meet others from different backgrounds who have different opinions, preferences, lifestyles, and ethnicities (Putnam 2000, p. 338). Through involvement in voluntary associations, individuals can learn more about minorities and the importance of tolerance and societal co-existence, which influence their attitudes. From this perspective, it can be assumed that intolerance develops as a reaction to the threat posed by out-groups. People may perceive individuals from minority groups as a threat to their value system, their economic or physical existence, or their social status, which in turn has consequences for the level of tolerance (Mutz 2002). However, people can learn to react to social diversity in tolerant and intolerant ways. In brief, associations are assumed to be "learning schools for democracy", fostering democratic attitudes and values such as tolerance and concern for other people (Warren 2001;Mutz 2002;Morales and Geurts 2007).
However, the socialization perspective has been both questioned and criticized. The argument is that positive correlations between membership and interpersonal attitudes may be the result of self-selection whereby individuals who hold tolerant views are more likely to become members in voluntary associations or work voluntarily than those who hold intolerant views (Uslaner 2002;Quintelier 2013;Van Ingen and Bekkers 2015). Indeed, many voluntary associations are homogeneous and segregated, and serve as meeting places for people with similar backgrounds (Baggetta 2016;Van der Meer 2016). To be with like-minded others may be the very reason individuals join voluntary associations or volunteer for a good cause.
Nevertheless, although self-selection may play a part in this argument, membership in voluntary associations and volunteer work may still be associated with a change in attitudes toward immigrants (e.g., Hooghe and Quintelier 2013). One argument is that voluntary associations differ in their potential both to generate democratic attitudes and to affect the development of attitudes towards immigrants (cf. Stolle and Rochon 1998;Warren 2001). The efficacy of voluntary associations in terms of civic values and attitudes is apparent in the distinction between bridging and bonding associations (Putnam et al. 1994, p. 90;Putnam 2000, p. 22). Bonding associations bring together people who are alike when it comes to central aspects such as gender, class, ethnicity, or political orientation. Bridging associations, on the other hand, bring together individuals who are unalike when it comes to such aspects. It is often argued that bridging associations have the most positive effects on the development of democratic attitudes and that they instil tolerance and acceptance of otherness (Paxton 2002).
Furthermore, group socialization theory stresses that when individuals become members of a particular group, they come to identify with that group and accept its "rules, standards, and beliefs about appropriate conduct and attitudes" (Turner 1987, p. 1). By associating with a voluntary association or becoming a volunteer, an individual may reinforce and/or internalize the values, norms, and identities that are constitutive within the group of members or supporters (Tajfel 1981). Over time, groups attempt to change the individual so that the individual member can contribute more to the explicit or implicit goals of the group (Levine and Moreland 1994, p. 309). The position they converge upon has normative consequences and leads to increased social influence that is relevant for the formation of norms and attitudes (Levine and Moreland 1994;Postmes et al. 2005). In the literature on voluntary associations, this has been referred to as the "cultural causal perspective" (Hooghe 2003) and suggests that if attitudes toward immigrants are significant and constitutive for identity in voluntary associations, such values and norms will be internalized by members and volunteers. As such, it is possible that existing attitudes of tolerance can develop further within associations or a group that has tolerant members and volunteers.
Research by Coffé and Geys (2007) shows that hobby clubs, humanitarian organizations, and art groups offer the best opportunity to bridge demographic divides between members. Similarly, Paxton (2002) found that peace and human rights associations offer the most substantial bridging potential. In addition, Hooghe and Quintelier (2013) found that membership in cultural associations, ethnic associations, and trade unions result in less intolerant attitudes, while sports associations had no such effect. Taken together, this would suggest that membership in certain types of voluntary associations may relate to the development of attitudes toward immigrants. It is also possible that involvement in voluntary associations and volunteer work over several years may have a greater effect on the development of attitudes toward immigrants compared to shorter-term involvement.

This study
Despite extensive research, the developmental effect of membership in voluntary associations and volunteer work on anti-immigrant attitudes remains unclear, in particular in respect to adolescents. This study used four waves of data as a means to address some of the limitations of previous studies. This was done in several ways. First, we adopted a longitudinal approach to examine how membership status (i.e., being a member or not) in various types of voluntary associations and volunteer work relate to anti-immigrant attitudes. Second, we examined how anti-immigrant attitudes develop and change over time. Third, we examined how the total number of years spent as a member or volunteer impact the change in attitudes that adolescents have toward immigrants. These aspects we refer to as the duration of membership and volunteer work. Finally, we examined the impact of the total number of uninterrupted years as a member in the same type of association. We refer to this aspect of membership as the persistence of membership.

Participants
To answer our research question, we used four waves of data from a longitudinal project. The data were collected annually and taken from a community-based project that was conducted in a medium-sized Swedish city with a total population of about 135,000. The reason for choosing this specific city relates to its representativeness in terms of national demographic characteristics. According to official statistics (Statistics Sweden 2012), the city in 2010-the first year in which data were collectedwas similar to the national average in terms of annual mean income (234,058 Swedish kronor/person compared with 237,186 Swedish kronor/person for the whole country), rate of unemployment (8.6% compared with 8.0% for the whole country), and percentage of foreign-born residents (14.8% compared with 14.7% for the whole country).
Thirteen schools within the city were selected for the study: 10 of its 14 junior high schools and 3 of its 7 senior high schools. The schools were strategically selected both on the basis of their different features, which meant that various types of schools (public and private) would be represented in the study, as well as on the basis of the diverse socio-economic and ethnic backgrounds of their students. The seventh graders in the junior high schools and the first-year students in the senior high schools were the target groups for the study. The target sample included approximately 2000 students at each time point. The participation rate ranged from between about 85 and 90% across the four time points. In the analytical sample, we only included respondents for whom we had data on anti-immigrant attitudes on at least two time points. The final analytical sample, then, included 1773 participants (50.7% girls; M age = 14.96 years and SD age = 1.70 at the first time point, i.e., 2010).

Attrition analyses
To establish whether drop-outs differed from those who participated in the study in terms of certain independent variables [i.e., gender, age, subjective socio-economic status (SES), immigrant status, membership in associations, volunteer work, etc.], we applied logistic regression analysis to all of these. None of the variables, in statistical terms, significantly predicted youth attrition, meaning that participants and drop-outs were not different, which indicates that attrition did not have an impact on the results.

Procedure
The respondents completed the questionnaires in their classrooms during regular school hours. To address the fact that the presence of teachers might influence the responses of students, we made sure that there were no teachers in the classrooms during data collection. Trained research assistants distributed the questionnaires and informed the students that participation in the study was voluntary; the students were also assured as to the confidentiality of their responses. No student was paid to participate in the study, but each class received a contribution of 1000 Swedish kronor (approximately US$ 125) toward its class fund. The data were collected annually from 2010 to 2013.

Anti-immigrant attitudes
The anti-immigrant attitudes of adolescents, here relating to negative and xenophobic attitudes toward immigrants, were measured using three items. The respondents were asked to respond to the following three statements about people who have fled to Sweden from other countries: "It's too often the case that immigrants have customs and traditions that do not fit into Swedish society"; "Immigrants often come here just to take advantage of welfare in Sweden"; and "Immigrants often take jobs from people who were born in Sweden". Respondents were asked to report on how true these statements were on a 4-point scale ranging from "1" (does not apply at all) to "4" (applies very well). In this study, Cronbach's α for this scale ranged from 0.75 to 0.80 across the four measurement points.

Membership in voluntary associations
With respect to membership in voluntary associations, adolescents were asked to position themselves on the following questions: "Are you a member of an association/associations?" The response options were "yes" and "no" with an opportunity to specify the type of association. The types of associations that the adolescents could select were as follows: sports association, cultural association, religious association, hobby association, recreational association, political association, association for peace and human rights, immigrant association, and environmental association. To take into consideration different aspects of adolescents' membership in associations, we created three measures, which are explained in the following paragraphs. The first variable, membership status, measures whether the adolescents are members in any associations or not at the first measurement point. The second variable, duration of membership, measures the total number of years in which youths claim to be a member of any association. This variable ranges from 0 to 4, where 0 indicates that the respondent has not been a member of any association at any measurement point, and 4 that the respondent reported membership in at least one association across the four measurement points. Finally, persistence of membership measures whether the respondent remained a member of the same type of association across time. More specifically, this variable is an additive index and measures the total number of uninterrupted years of membership in the same type of association across time. As four types of associations are included in the creation of membership measures at four measurement points, the response scale for this specific measure can supposedly range from 0 to 16. For example, 0 means the respondent has not been a member of any association at any time, 1 indicates membership in at least one association at one time point, 2 indicates the respondent has been a member of the same type of association for two years in a row, and so forth.

Volunteer work
We also included a measurement of volunteer work. In the questionnaire, the adolescents were asked to take a stand on how often in the past 12 months they had worked voluntarily for a good cause. The response options were "yes, once or twice," "yes, a number of times", and "no". This single item was recoded into "0" (have not worked voluntarily) and "1" (have worked voluntarily). Based on this item, we created a second measure, duration of volunteer work, to take into consideration the possible impact of the longevity of adolescents' volunteer engagement. More specifically, this variable measures the total number of years (0-4) of volunteer work.

Socio-demographic characteristics
To consider the effects of some other potentially important factors, we included several conventionally used control variables: age, gender, immigrant status, and subjective socio-economic status (SES). Gender was coded as "0" (boy) and "1" (girl). Immigrant status was measured as a dichotomous variable (0 = immigrant-the adolescent and both parents were born outside Sweden; and 1 = Swedish-the adolescent and at least one parent was born in Sweden). More specifically, if both one parent and the respondent were born in Sweden and the other parent outside Sweden, the respondent was coded "Swedish". If the respondent was born in Sweden, but both parents were born outside Sweden, then the respondent was coded "Immigrant". Subjective SES was measured using the following five questions: (1) "If you want things that cost a lot of money (e.g., a computer, skateboard, cell phone), can your parents afford to buy them if you want them?", (2) "If you compare yourself with others in your class, do you have more or less money to buy things?", (3) "Does your family have more or less money than other families where you live?", (4) "How often do you and your family go on vacation?", and (5) "What are your family finances like?" Cronbach's α reliability for this scale was 0.79 at time point 1, that being the year 2010.

Research strategy
To examine the stability and explain the development of or changes in antiimmigrant attitudes among adolescents over time, a series of statistical analyses were conducted. First, to examine the stability of anti-immigrant attitudes longitudinally, we analysed three measures of correlation: Pearson's product moment correlation; Kendall's τ-b; and Spearman's ρ rank correlations. In so doing, we were able to examine the relative stability, that is, the extent to which adolescents change their attitudes relative to one another over time. Second, to examine the role of voluntary associations so as to find an explanation for the development of, or changes in, anti-immigrant attitudes among adolescents over time, we constructed a series of latent growth curve models (LGCMs; McArdle and Nesselroade 1994; Geiser 2013), using maximum likelihood estimation in Mplus 8 (Muthén andMuthén 1998-2017). These models, also known as latent curve analysis models, generate optimal means for studying changes over time. With longitudinal data, they focus mainly on patterns of growth and decline. More specifically, in this study, growth and decline in anti-immigrant attitudes were considered a function of a latent intercept and a latent slope. The latent intercept reflects the average initial value of anti-immigrant attitudes at the start of the longitudinal change process. Put differently, the intercept shows the mean of antiimmigrant attitudes at the first time point. The latent slope indicates the average individual change rate (over the four measurement occasions). A positive mean value for the slope indicates that, on average, there is an increase in the variable over time, whereas a negative value indicates a decline. For this study, taken as a whole, latent growth curve analysis was a suitable statistical procedure for investigating whether predictors from the voluntary associations can explain the rate of change in anti-immigrant attitudes over time.
With regard to the latent growth curve analyses, two statistical clarifications are required. First, to evaluate the models statistically, we used the following goodness-of-fit indices (cf. Kline 2010): χ 2 , the Comparative Fit Index (CFI), the Root Mean Square Error of Approximation (RMSEA), and the Standardized Root Mean Squared Residual (SRMR). Generally speaking, these fit indices are estimates of how well a theoretical model matches observed data. Thus, a good model fit implies that the data are consistent with the assumptions of a hypothesized model. According to recommendations, a good fit to data is indicated by a non-significant χ 2 , CFI equal to or higher than 0.95, RMSEA equal to or lower than 0.05, and SRMR equal to or lower than 0.08 (Hu and Bentler 1999).
Second, to deal with missing data, we tested all our models using Full Information Maximum Likelihood (FIML) estimation. By making use of all available data from every subject in a sample, FIML estimation provides more reliable standard errors than listwise deletion, pairwise deletion, or mean imputation (Little and Rubin 2002). The proportion of missing data values is examined by the covariance 'coverage' matrix in Mplus. The minimum acceptable covariance coverage value, for which Mplus provides a robust estimate, is as low as 10% (Muthén andMuthén 1998-2017). In this study, coverage was much higher, ranging from 56 to 88%.
Finally, to account for the multilevel structure of our data (adolescents nested in 13 schools), we used the Mplus option "type = complex", with schools treated as a clustering variable. This procedure corrects for possible bias in standard errors due to the multilevel structure of the data (McNeish et al. 2017;Muthén andMuthén 1998-2017).

Membership in various types of voluntary associations
In this section, the results from the empirical analyses are presented. Before studying the role membership in an association plays in the anti-immigrant attitudes of adolescents, we took a closer look at the types of associations that adolescents choose to become members in. As shown in Table 1, three types of associations attract a particularly large share of young people, namely sports associations, cultural associations, and religious associations. More than half of the young people (T1, 2010) are members of sports associations, approximately one-fifth are members of cultural associations, and more than one-tenth are members of religious associations. However, a significantly lower proportion of adolescents are members of, for example, outdoor recreational associations, immigrant associations, political associations, environmental associations, and associations for peace or human rights. In sum, young people's engagement varies depending on the type of association.
In addition, results in Table 1 show that for most associations, adolescents' involvement tends to decline with age. This trend is most apparent with sports associations in which more than 51% of the adolescents were members during the first year (T1) compared with 38% at the fourth measurement point (T4). A similar trend can be noted with, for example, cultural associations, in which nearly 22% of the young people were members at T1 (2010) but less than 15% were 3 years later (T4). However, certain types of associations show a moderately opposite trend in terms of attracting young people. For example, about 12% of adolescents (T1) were members of religious associations, while the corresponding figure for T4 was just over 13%. A similar trend is noticeable for political associations. In other words, adolescents are attracted by somewhat different types of associations depending on their age. Furthermore, our analyses also show the percentage of adolescents who stated they had worked voluntarily for a good cause in the past 12 months. Approximately every second respondent stated that they had done volunteer work. However, involvement in volunteer work seems to decrease slightly with age, from 52.6% at T1 to 46.8% at T4.

Differences between member and non-member
To answer the question of whether membership in certain associations can explain the development of attitudes toward immigrants over time, we first need to establish whether there are any differences at all in the level of anti-immigrant attitudes between members and non-members. To this end, a series of t-test statistics were conducted, and effect sizes (Cohen's d) were calculated for the differences between members and non-members in terms of anti-immigrant attitudes. The magnitude of the difference was considered small (0.2), moderate (0.5), or large (0.8).
The results of the mean difference tests are presented in Table 2, which shows the mean values of anti-immigrant attitudes for members and non-members broken down by type of association. To begin with, we can state that there are statistically established differences between members and non-members in several associations. Statistically established differences are indicated with figures in bold.
As shown in Table 2, members of cultural and religious associations express less anti-immigrant attitudes at all four measurement points. A similar pattern is apparent for adolescents who are members of political associations and for adolescents who are members of peace and human rights associations. However, two types of associations show opposite results. Hobby associations and sports associations stand out in that members express slightly higher anti-immigrant attitudes compared to nonmembers. At the same time, no significant mean differences were found between members and non-members in the three remaining types of associations, that is, outdoor recreational, immigrant, and environmental associations. Finally, we can also note that adolescents who have engaged in volunteer work once or twice express less anti-immigrant attitudes than those who have not engaged in similar activities. In general, the effect sizes were small to moderate (Cohen's d ≤ 0.20 to < 0.80).
To conclude, adolescents who are members of certain types of associations, particularly cultural, religious, peace/human rights, and political associations, express somewhat less anti-immigrant attitudes than those who are not members. Moreover, adolescents who have engaged in volunteer work are somewhat less anti-immigrant in their attitudes than those who have not. Based on these results, we will in all forthcoming analyses focus only on the four associations in which members had statistically lower levels of anti-immigrant attitudes.

Stability of anti-immigrant attitudes over time
After examining the differences between members and non-members in terms of their level of anti-immigrant attitudes, we conducted a series of correlations to examine the stability coefficients-that is, the correlation between anti-immigrant attitudes at two measurement occasions, for example T1 and T2 or T2 and T3. The results from these analyses are presented in Table 3. As can be seen, adolescents' anti-immigrant attitudes seem to stabilize with age. For example, the stability coefficient between T2 and T3 (r = 0.620) is higher than the correlation between T1 and T2 (r = 0.521). When comparing the two correlation coefficients (i.e., 0.620 and 0.521), we found a statistical difference between the two correlations (z = − 3.81, p < 0.001), indicating that the level of stability between T1 and T2 is significantly lower than between T2 and T3. Moreover, Pearsonʼs r between T1 and T2 was also statistically different from the correlation between T3 and T4 (z = − 4.74, p < 0.001). The increase in stability over time is also reflected in the other two measures of correlations, that is Kendallʼs τ-b and Spearmanʼs. All in all, regardless of the type of correlation measure used to examine the stability pattern of anti-immigrant attitudes across time, the results showed a greater degree of instability at a younger age and an increase in stability as the adolescents increased in age.

Evaluating change in anti-immigrant attitudes over time
After addressing the question of when anti-immigrant attitudes stabilize, we proceeded to analyse the direction of change in variables of interest over time. To this end, we conducted a latent growth curve model (LGCM), where measures of antiimmigrant attitudes at four measurement occasions were included as manifest variables and served as indicators of two latent factors: the intercept, which, as mentioned before, reflects the mean level of anti-immigrant attitudes at the first data occasion; and the slope, that is, the amount of linear change occurring between the four time points. This model yielded a good model fit, χ 2 (5) = 37.78, p < 0.001, CFI = 0.96, RMSEA = 0.044, p = 0.143, 90% CI 0.04-0.08, SRMR = 0.034. The results from the LGCM for anti-immigrant attitudes are reported in Table 4. The mean intercept of anti-immigrant attitudes shows that the young people's mean value for anti-immigrant attitudes starts at 2.23, and this mean value decreases somewhat (β = − 0.01) between the first and the fourth measurement points (see Fig. 1). However, the decline in anti-immigrant attitudes over time is not statistically significant. At the same time, the variance estimates pointed to considerable inter-individual differences, as reflected in the variance of intercept and slope. The inter-individual variability in anti-immigrant attitudes over time is also depicted in Fig. 1. As can be seen, the growth curves of 50 randomly sampled respondents show substantial difference over time. While some individuals show a decline in antiimmigrant attitudes, others show an increase or remain stable over time.
To conclude, the significant variances of the intercept and slope, as well as the reasonable variability (as shown by the variance estimates), indicate that individuals differed in their anti-immigrant attitude scores at the first time point and also in their growth trajectories (see Fig. 1). All in all, these findings indicate overall stability in anti-immigrant attitudes over the four measurement occasions but also reasonable inter-individual variability.
Given the significant intercept and slope variances, there was justification for adding variables to the models to explain the variation in individual trajectories of anti-immigrant attitudes. To this end, we estimated a series of conditional latent  Table 5 Latent growth curve models of anti-immigrant attitudes Entries for covariates are standardized maximum likelihood estimates and standard errors *p < 0.05; **p < 0.01; ***p < 0.001 a The models are controlled for immigrant status, sex, SES, age, and the multilevel structure of the data The first four association types in Table 2 Single model − 0.14*** growth curve models, where each predictor was initially added separately to examine its predictive value, while not considering other influences. The results are presented in Table 5.
Looking at the results, we can see that membership in at least one association at T1 (2010), duration of membership, and persistence of membership, all related significantly and negatively to anti-immigrant attitudes at the starting point (intercept). Put differently, being a member of an association related to lower levels of anti-immigrant attitudes at the starting point (T1, 2010). Duration of membership (β = − 0.21, p < 0.001), that is the total number of years of membership in any of the four selected associations, was most strongly associated with adolescents' antiimmigrant attitudes at the starting level. Moreover, in a check of the various sociodemographic variables and the multilevel structure of the data, the results show that both duration (β = − 0.09, p < 0.05) and persistence of membership (β = − 0.08, p < 0.05) had a statistically significant impact on the change (slope) in anti-immigrant attitudes during the period that was studied. This means that the anti-immigrant attitudes of adolescents who were members of at least one association over a long period of time or who remained in the same association for a number of years in a row decreased more than those of adolescents who were not members at all or who did not remain members for any significant time. Furthermore, the results show that the anti-immigrant attitudes of adolescents who worked voluntarily (β = − 0.14, p < 0.001) decreased more than the anti-immigrant attitudes of those who did not work voluntarily.
Additionally, because members in sports associations showed higher levels of anti-immigrant attitudes than non-members, we conducted separate analyses for this type of association. The results showed that membership of sports associations was not significantly associated with the anti-immigrant attitudes at T1 nor with the change in them over time (slope), even after a check was conducted of different demographic factors.
To conclude, the results of our analyses show that several aspects of membership in associations and volunteer work are associated with the starting level of and change in anti-immigrant attitudes. In addition, the impact of engagement in volunteering on any change in anti-immigrant attitudes seems to be slightly greater than the impact of duration and persistency of membership.

Conclusion and discussion
The aim of this article is to contribute to the understanding of the role that membership in voluntary associations and volunteer work play in the development of anti-immigrant attitudes among adolescents. More precisely, we posed the following research question: to what extent do membership in voluntary associations and volunteer work influence the development of anti-immigrant attitudes among adolescents? The question was answered by using a unique four-wave longitudinal data set from an extensive Swedish research project about young people's socialization. The target sample comprised approximately 2000 adolescents between the ages of 13 and 16 at the first measurement occasion. In our analyses, we took into consideration the impact of membership in various types of associations and dimensions of membership and volunteer work.
The most important finding of this article is that long-term membership in voluntary associations and long-term volunteer work reduce anti-immigrant attitudes among adolescents. In other words, the results show that the anti-immigrant attitudes of adolescents who were members of at least one association over a long period of time or who remained in the same association for several years in a row decreased more than those of adolescents who were not members at all or who did not remain members for long. Furthermore, we found that the anti-immigrant attitudes of adolescents who worked as volunteers decreased more than the anti-immigrant attitudes of those who did not work as volunteers. However, the relationship with respect to membership is less strong statistically than volunteer work and holds true only for adolescents who are members in certain types of associations, namely cultural and religious associations, peace and human rights associations, and political associations.
Taken together, the results of this article provide empirical support for the general argument that civic engagement has positive benefits on the generation and stimulation of interpersonal attitudes (Putnam et al. 1994;Putnam 2000;Warren 2001), at least with regards to adolescents' attitudes toward immigrants. In other words, the results suggest that voluntary associations and volunteering may indeed be relevant for reducing prejudice toward immigrants. However, the results indicate that not all voluntary associations have the same potential to affect adolescents' interpersonal attitudes as suggested in the literature (Putnam 2000). Instead, voluntary associations are able, in different ways, to influence the development of and change in adolescents' attitudes toward immigrants. As such, our findings are in line with the study by Quintelier and Hooghe (2013), who found that ethnocentrism was lower among members of ethnic groups, religious groups, political parties, and new social movements. However, this article adds empirical insight into the fact that long-term volunteering has an equally great (or even greater) impact on the development of and change in adolescents' attitudes toward immigrants as does membership in certain type of associations.
In this article, we have not been able to account for the exact mechanism through which membership in voluntary associations and volunteer work relate to attitudes toward immigrants. Thus, it is difficult to determine the extent to which the results presented here are caused by socialization whereby individuals learn the importance of tolerance and societal coexistence as suggested by Putnam (2000) or an internalization of the values, norms, and identities constitutive to certain types of associations as suggested by the cultural causal perspective (Hooghe 2003). To fully understand the casual mechanism, one would need to take into consideration, for example, the activities taking place in the associations and network of volunteers, and the norms, values, and ethnic composition of the volunteer associations and networks that adolescents choose to join. Ideally, one would also need to check the time the young people spend volunteering and being members in the various types of voluntary associations, as well as how they demonstrate their civic engagement.
However, the finding that membership in specific types of voluntary associations influences the development of anti-immigrant attitudes indicates that certain types of values, norms, and identities conductive to positive interethnic attitudes are promoted in these types of associations. It could be argued that values and norms such as tolerance, respect, humanitarianism, and peaceful coexistence are more strongly communicated in religious associations, peace and human rights associations, political and cultural associations than they are in, for example, sports associations, and hobby associations and outdoor recreational associations. Similarly, at an individual level, volunteering is often associated with values and attributes such as altruism, solidarity, reciprocity, beneficence, and equality. In addition, participating in volunteer work signals a pro-social behavior and a commitment to working with others in the interests of the public good (Dekker and Halman 2003). As such, it is possible that the internalization of such values and norms also reduces the level of anti-immigrant attitudes of members in networks of volunteers. What must be considered, however, is that we cannot check the effects of self-selection. Thus, it is possible that the long-term effect is valid preferably for adolescents already committed to the innate norms and values of these types of associations and networks when they first became involved in them. In other words, membership in voluntary associations and volunteer work to reinforce tolerance attitudes among individuals who are likely to accept norms of tolerance/already tolerant individuals.
This study has both limitations and strengths that warrant attention. A first limitation is that we do not consider the exact mechanism through which membership in voluntary associations and volunteer work is supposed to relate to attitudes toward immigrants. Second, we cannot account for all other factors that may play a role in influencing anti-immigrant attitudes such as peers, parents, and schools (cf. Jones and Rutland 2018). Another limitation is that we do not account for the effects of self-selection. To establish with certainty that membership in an association or volunteer work has an impact on attitudes of intolerance, we need to consider the individual's level of anti-immigrant attitudes before he or she becomes active. It is important to note that the young people who were already members of a voluntary association or who were working as volunteers at the start of the study expressed very low levels of anti-immigrant attitudes, which means there was limited scope for them to decrease their level of intolerance, and thus there are only small differences to explain. This may be due to self-selection, that is, young people with lower anti-immigrant attitudes who choose to volunteer and become members of associations. A fourth limitation of this study that we want to point out is that we cannot account for the time spent in voluntary associations and volunteering. However, using a longitudinal approach following the same individuals over time, we can explore to what extent, if any, civic engagement relates to the development of certain attitudes toward immigrants.
Our study has, at the same time, several strengths. First and foremost, by adopting a four-wave longitudinal approach, we were able to examine how membership in voluntary associations and volunteer work influence the development of anti-immigrant attitudes among adolescents. In so doing, we were able to address the long-term impact of civic engagement and to get beyond the crosssectional correlations from previous research. A second strength concerns the focus on adolescents. As noted previously, research in developmental psychology as well as in political science has demonstrated that adolescence is a critical period in a person's life when attitudes are open for change. Therefore, adolescence seems to be a good starting point for an exploration of different aspects of civic engagement relating to anti-immigrant attitudes. Taken as a whole, despite its limitations, this study presents important insights into previous research concerning the role of civic engagement in the development of attitudes. Among other things, it provides evidence that being a member of a voluntary association and engaging in volunteer work over a long period contribute to reducing antiimmigrant attitudes among adolescents. In other words, time matters.