Social Differences in Youth Volunteering in an Increasingly Diverse Society: An Intersectional Approach

Migration Background

The first social factor examined in this article is migration background. Although research indicates that formal volunteering can help bridge cultural divides between individuals with and without a migration background, people with a migration background are often underrepresented in formal volunteering (Siongers & Spruyt, 2025). This underrepresentation can follow from two different mechanisms.

First, the act of migration itself can negatively affect formal voluntary engagement (Ambrosini & Artero, 2023; Musick & Wilson, 2008). First-generation migrants often have smaller social networks, as they leave behind established connections and require time to build new ones in the host country (Ambrosini & Artero, 2023; Southby & South, 2016). A smaller social network reduces the likelihood of being introduced to individuals already active in volunteering and, as such, also to voluntary work and organizations. In addition to smaller social networks, first-generation migrants also face various barriers to volunteering, such as language difficulties and limited familiarity with the cultural norms that shape organizational life (Cappelletti & Valtolina, 2015).

A second strand of literature argues that sociocultural differences and the deprivation or discrimination that people with a migration background face hinder their voluntary engagement. This applies not only to first-generation migrants but also to second—and possibly further—generations (Hunkler et al., 2015). A crucial finding here is that many individuals begin volunteering because they are invited to do so, rather than initiating it themselves (Bekkers et al., 2016). People with a migration background are often not regarded as perfect volunteers and are therefore less frequently asked to volunteer (Southby & South, 2016; Sundeen et al., 2007). These socio-ethnic differences in being recruited are also likely to vary by the specific country of origin, given that discrimination (one but not the only cause) has been shown to vary by the country of origin, with often a European/non-European divide in discrimination in European countries (Acolin et al., 2016; Ghekiere et al., 2023). Moreover, taste-based discrimination also often differs in terms of countries of origin, with those with a European background often facing less discrimination than those with a non-European background (Ghekiere et al., 2023).

Additionally, the culture of a migrant’s home country can affect their attitudes towards volunteering (Randle & Dolnicar, 2009). Cultural norms may influence the willingness to participate, with different ethnocultural groups viewing volunteering and helping others in unique ways (Falzarano et al., 2022; Randle & Dolnicar, 2009). Hofstede (1980) differentiates between individualistic and collectivist societies, noting that individualistic cultures may value volunteering less. In Belgium, the individualism difference between Moroccan (more individualistic) and Turkish (more collectivist) communities is well documented (Lesthaeghe & Surkyn, 1997).

Besides general cultural differences, country-specific variations in volunteering rates exist among Belgium’s largest migrant groups from Türkiye, Morocco, Poland, Congo and Mediterranean countries. Research by Aslan and Tuncay (2024) indicates low volunteering rates in Türkiye, with limited national promotion. Belgium’s neighbours have similar voluntary engagement levels, suggesting comparable attitudes (Enjolras, 2021). Poland and Mediterranean countries rank low, indicating hesitancy towards formal volunteering (Charities Aid Foundation, 2023; Nakamura et al., 2025). No current data are available for Morocco and Congo (UN Volunteers, 2026).

To sum up, we expect differences not only by young people’s migration status but also by the country of origin/background. Hesitant or adversarial attitudes towards volunteering in one culture might thus lead to lower volunteering rates among migrants from that culture in another nation that is more open to volunteering.

Gender

There is no consensus on gender differences in (formal) voluntary engagement. Some studies suggest an overrepresentation of men, while others indicate an overrepresentation of women (Musick & Wilson, 2008; Wymer, 2012), and yet others find no gender differences (Siongers & Spruyt, 2025). Gender differences are more likely to be found in specific contexts of voluntary engagement (Einolf, 2011). For example, research shows that in some organizations, such as fire stations, men are overrepresented, while women are more likely to volunteer in organizations involved in care tasks (Einolf, 2011; Musick & Wilson, 2008). Gender differences are also found within organizations based on the division of labour: men are often assigned technical or prestigious roles, whereas women are more frequently allocated caregiving and administrative tasks (Rotolo & Wilson, 2007).

Various explanations have been proposed for these differences, in access to both volunteering and task distribution. A first explanation refers to the strong link between altruism, volunteering and charitable giving, with women consistently scoring higher on altruism and generosity (Musick & Wilson, 2008; Simmons & Emanuele, 2007). Among young people, girls tend to embrace more prosocial values and do so at a younger age, which may lead to higher volunteer participation (Beutel & Johnson, 2004).

Second, women generally have closer and stronger social networks than men (Blyth & Foster-Clark, 1987). Since being invited is a key factor in volunteering (Bekkers et al., 2016), stronger networks increase the likelihood that (young) women are asked to participate and encouraged to do so (Wymer, 2012).

A third relevant explanation relates to household and care tasks. Women and girls continue to perform a disproportionate share of household work (Herd & Harrington Meyer, 2002; Mullens & Glorieux, 2019; Putnick & Bornstein, 2016). The implications for volunteering are twofold. On the one hand, according to the time-scarcity hypothesis, increased domestic responsibilities reduce the time available for volunteering (Herd & Harrington Meyer, 2002; Mullens & Te Braak, 2023). On the other hand, household and care responsibilities can lead to increased voluntary engagement. Gendered socialization may promote care-taking behaviour in girls, encouraging both informal and formal voluntary engagement (Musick & Wilson, 2008; Simmons & Emanuele, 2007; Wuthnow, 1995).

Taken together, these divergent mechanisms may cancel each other out, which could explain why some studies find no significant gender differences in volunteering.

Social Status

In this article, social status refers to the socio-economic position of young people, reflecting their standing in society, and is defined in broad terms. We use the term as having two components: a material and a cultural component. For young people, social status is measured by their parents’ income (material component) and their educational track as a cultural component. While parental income is an ascribed characteristic beyond the young people’s control, the educational track is considered a (partly) acquired one. Despite persistent patterns of social reproduction in Flanders (and hence a strong link between parents’ education and young people’s track position; see Elchardus et al., 2013), it is ultimately the young person who navigates their educational trajectory (Lagaert et al., 2024).

Educational attainment is one of the strongest predictors of participation in voluntary work (Boraas, 2003; Musick & Wilson, 2008). Differences based on education are often attributed to social networks and to being ‘asked’ to volunteer, two factors already discussed. According to the so-called relative education model, it is not the specific knowledge and skills acquired that explain differences in voluntary engagement among young people, but rather the social-network position education affords. Organizations tend to recruit volunteers from individuals who occupy central positions within social networks. Educational status plays a key role in securing that centrality (Persson, 2011), and this especially applies in countries with a highly tracked secondary education, such as Flanders. Research shows, for example, that in Flanders, young people’s social networks (including friendships) are strongly segregated by educational tracks and that young people who follow vocational education experience track-based stigma (Spruyt & Van Droogenbroeck, 2024). This social status not only increases the likelihood of being asked but is also linked to what Lareau, following Bourdieu, calls a ‘sense of entitlement’ (Golann & Darling-Aduana, 2020; Lareau, 2011). Young people from higher socio-economic backgrounds are more likely to pursue general secondary education, which not only confers higher status but also offers access to broader networks and greater institutional experience. In this way, the educational track serves as a key channel through which social status translates into differences in voluntary engagement.

In terms of material resources, income may also influence volunteering. The dominant status theory holds that dominant groups in society are more likely to volunteer. Dominant groups often possess more economic, social and cultural resources, which can be attributed to volunteer work (Hustinx et al., 2022; Musick & Wilson, 2008). Furthermore, these groups are seen by volunteering organizations as more desirable than those with fewer economic, social and cultural resources (Hustinx et al., 2022). This means that the economic status often provides access or barriers to entry into volunteering organizations.

A second way the economic aspect of social status might affect young people’s odds of volunteering is through the time-scarcity hypothesis. With part-time work becoming more prevalent among Belgian youth, there may be less time available for voluntary activities (Spruyt, 2024). Furthermore, part-time work is more prevalent in low-income households (Spruyt, 2024), suggesting that young people in low-income households might spend more time on income-generating activities than on unpaid, voluntary activities.

Intersectionality

In this article, we not only focus on the relative importance of each characteristic. We are primarily interested in assessing the relevance of the combinations of these characteristics. Crenshaw (1991) first introduced the concept of intersectionality to describe how multiple identities intersect and create new forms of discrimination or disadvantage.

Since its introduction, the concept of intersectionality has been further developed and broadened and is no longer limited to gender and ethnicity. Intersectionality can be understood as the intersection of different forms of inequality and exclusion, such as racism, sexism and class discrimination (Combahee River Collective, 1977; Wekker & Lutz, 2001). The central idea is that different axes of inequality also lead to interlinked systems of oppression and exclusion (Collins, 1993). In this contribution, we adopt an intersectional perspective to examine access to and participation in volunteering.

There is growing evidence that an intersectional approach provides valuable insights into inequalities in voluntary engagement. Many studies have shown that the intersection of different identities creates unique positions regarding volunteering. For example, Musick and Wilson (2008) found that higher income promotes volunteering among white Americans, but not for black Americans. Musick and Wilson (2008) show that status differences are often reinforced by people’s ethnicity. One might even expect that both characteristics (i.e., migration background and social status) reinforce one another as volunteer organizations frequently continue to associate the ideal volunteer with individuals who possess high levels of social capital, such as the highly educated and native-born. Those who deviate from this profile face reduced opportunities to engage in voluntary work.

Beutel and Johnson (2004) found a difference in prosocial values, an important predictor of volunteering, between white boys and girls, but not between black boys and girls. This finding suggests that gender differences may become salient in interaction with other social factors, such as ethnicity. Research has shown that the effect of gender on volunteering varies by respondents’ ethnocultural background in the United States (Bellido et al., 2021) and that gendered participation in leisure organizations varies across ethnocultural backgrounds (Schaillée & Theeboom, 2014).

With this in mind, this article aims to supplement existing research in two ways. First, it highlights the need for more data on how identity characteristics interact in accessing volunteering, supporting an intersectional perspective. Second, while most intersectionality research relies on qualitative methods, this study deliberately adopts a quantitative approach. Unlike most qualitative intersectionality research, this study uses a quantitative approach to demonstrate that the interplay of characteristics should be evident from large-scale, representative data. At its core, the intersectional approach rests on the idea that the ‘effects’ of separate characteristics cannot be isolated from those of other characteristics. Translated in quantitative terms, this should mean that one finds at least some meaningful significant interaction terms, as these imply that the effect of one characteristic moderates the effect of another.

Summary and Expectations

Drawing on previous literature, we expect to find (a) social differences in young people’s participation in formal volunteering and (b) that characteristics such as gender, migration background and social status will reinforce or weaken each other’s effects.

Literature indicates that social factors, including migration background, influence young people’s volunteering. Migration-related factors, such as limited social networks and intergroup dynamics, including prejudice, are expected to negatively affect formal volunteering among youth (Ambrosini & Artero, 2023; Hunkler et al., 2015; Musick & Wilson, 2008; Sundeen et al., 2007). These effects are anticipated to vary across ethnic–cultural groups due to differing migration histories and perceived discrimination levels (Ambrosini & Artero, 2023; Hofstede, 1980; Musick & Wilson, 2008; Randle & Dolnicar, 2009). We therefore expect that having a migration background negatively affects young people’s odds of formal volunteering, with variation across countries of origin. Research shows a clear European/non-European distinction in discrimination at institutional entry points (Acolin et al., 2016; Ghekiere et al., 2023), suggesting that European respondents may face fewer negative effects than those from Maghrebi, Middle Eastern or Sub-Saharan backgrounds. Additionally, volunteering rates are expected to influence these effects; youth from countries with rates similar to Belgium’s may not experience negative effects, while those from countries with lower rates might.

Regarding gender, we do not expect a general gender difference in participation, as counteracting mechanisms (e.g., prosocial attitudes vs time constraints) may cancel each other out, consistent with earlier findings (Einolf, 2011; Musick & Wilson, 2008; Siongers & Spruyt, 2025; Wuthnow, 1995).

We further expect young people with a higher social status to volunteer more than those with a lower social status. Because the literature holds that those in more privileged social positions, based on both education and income, will be regarded more as the ‘ideal volunteer’ by organizations, they will have higher odds of formal volunteering than those in less-privileged social positions (Boraas, 2003; Golann & Darling-Aduana, 2020; Lareau, 2011; Musick & Wilson, 2008).

Furthermore, we expect that the intersection of migration background, social status and gender will also affect young people’s likelihood of engaging in formal volunteering. Previous research suggests that status effects may be intensified by migration background (Musick & Wilson, 2008) and that gender might constitute a double jeopardy for those with other discriminated identities (Beutel & Johnson, 2004; Musick & Wilson, 2008).

Based on the dominant status theory, we expect that young people in an underprivileged social position will experience negative effects on their chances of volunteering (Hustinx et al., 2022; Musick & Wilson, 2008). This suggests that young people with lower social status and those with a non-European migration background will face a double jeopardy for their precarious identities, which will decrease their volunteering. Furthermore, gender effects may interact with migration background, as cultural background influences gender disparities in volunteering (Bellido et al., 2021). Because this research was conducted in the American context, we find it prudent not to expect the same outcomes. We do, however, expect an interaction effect between gender and migration background, but cannot, in good conscience, predict its exact direction.

Data

To answer our research questions, we use data from the JOP Urban Youth Survey collected in 2023 among pupils in Flemish secondary schools. The survey placed specific focus on schools in the urban areas of Antwerp, Ghent and Brussels, three of the region’s most urbanized and diverse cities in Flanders (the Dutch-speaking part of Belgium), with a control group for the rest of Flanders. Data collection was conducted digitally, using tablets provided in participating schools. A total of 5,550 young people participated in the survey (2,875 girls and 2,675 boys) with a school-level response rate of 28.1%. To ensure the data set’s representativeness, post-stratification weights were applied (Lagaert et al., 2024). These weights were calculated based on population statistics for the 2022–2023 school year gathered by the Flemish Department of Education (October census). Both the urban sample and the control group for the rest of Flanders were weighted separately by gender, grade and educational track. Greater or lesser weight was assigned to the responses of certain individuals, depending on the under- or overrepresentation of that group in our sample relative to the proportions in society (Lagaert et al., 2024). In the multilevel analysis, post-stratification weights were applied at the individual level to ensure that parameter estimates accurately reflect the population distribution and to account for clustering within schools. Urban oversampling increases the precision of estimates for students from metropolitan areas without biasing associations after applying weights. This allows weighted results to be validly generalized to all students in Flanders, both urban and non-urban.

Two additional selections were applied. First, only respondents aged 15 or older were included in analyses related to formal volunteering, as this is the minimum legal age for such activities in Belgium. Second, to avoid treating young people with a migration background as a homogeneous group, we disaggregated this group based on the country or region of origin. To ensure sufficient statistical power and meaningful interpretation, only countries/regions with adequate sample sizes were retained in the analysis.

After applying these selections, the final sample comprises 2,354 young people (1,268 girls and 1,086 boys; mean age = 16.4).

Measures

Socio-demographic Variables

We classify young people as having a migration background if they or at least one of their parents were not born in Belgium (Centraal Bureau voor de Statistiek, 2025). The country of origin assigned to young people was determined by their mother’s nationality at birth. If the mother held Belgian nationality at birth and the father held a different nationality, the father’s nationality was used to determine the migration background.

To focus on the main migrant groups in Belgium and to ensure sufficient statistical power and meaningful interpretation, this analysis includes only young people from neighbouring countries (the Netherlands, France, Luxembourg), Mediterranean countries (Italy, Spain, Portugal), Congo, Türkiye, Morocco and Poland. These countries were chosen based on both the size of the respondent group in our data and the theoretical consideration that they are the most important countries of origin for migrants to Belgium over the last few decades (Statistiek Vlaanderen, 2025). Young people (n = 781) from other countries were excluded due to heterogeneity and limited sample sizes within this residual category. As a result, our sample consisted of 66.3% young people without a migration background (n = 1,561). In addition, 4.9% came from neighbouring countries (n = 115), 6.6% had a Turkish background (n = 156) and 15.3% had a Moroccan background (n = 360). Young people with Congolese (2.6%; n = 60), Polish (2.1%; n = 49) and Mediterranean (2.2%; n = 52) backgrounds were a smaller group within our sample.

Because the number of respondents from specific countries of origin was limited, resulting in a lack of statistical power, we constructed two versions of the migration background variable. First, migration background was operationalized as a categorical variable distinguishing between young people with and without a migration background and those with a European or non-European background. Second, it was operationalized as a categorical variable distinguishing between young people based on the country of origin (seven categories).

Gender was included as a dichotomous variable (46.1% boys, 53.9% girls). The 36 respondents who did not identify with either category were excluded from the analysis, as the group was too small to make reliable statements.

Social status was operationalized as a combination of young people’s educational position and their family’s subjective income. Subjective income was measured by the extent to which young people felt their families could make ends meet with their available monthly income. The five response categories ranged from ‘very difficult’ to ‘very easy’. If young people indicated that their family could get by easily or very easily, they were categorized as the high-income group. If they indicated they could get by reasonably easily, or if it was hard for them to get by, they were considered part of the low- to medium-income group.

Educational position was based on the educational track in which young people were enrolled. In the Flemish education system, there are three main educational tracks: general, technical and vocational. General education prepares students for higher education. Vocational education prepares students for entry into the labour market. Technical education provides access to both pathways. These different educational pathways differ not only in their finality but also in their social prestige (Spruyt & Van Droogenbroeck, 2024), with vocational education generally considered less prestigious. Moreover, the social reproduction of educational inequality is high due to the early tracking system (Dekeyser, 2024). In practical terms, this means that there is strong relationship between parents’ educational position (often seen as an indicator of cultural capital) and young people’s track position.

Social status was determined by combining young people’s educational track and subjective income and categorizing them into four groups: high, mid-high, mid-low and low. Respondents were categorized as having a high social status when they followed a general education and had a high income. If they followed a general track and had a low to medium income, or followed a technical track and had a high income, they were categorized as having a mid-high social status. Young people with a mid-low social status were those following a technical track without a high income and those following a vocational track with a high income. Young people in vocational education with a low to medium income were assigned a low social status. Table 1 shows the frequencies of all variables used in our study.

OPEN IN VIEWER

Table 1.

Frequency Table for all Variables Used.

		N	%
Gender
	Boys	1,086	46.1
	Girls	1,268	53.9
Social status
	High social status	909	38.6
	Mid-high social status	579	24.6
	Mid-low social status	611	26
	Low social status	255	10.8
Migration background
	No migration background	1,561	66.3
	EU migration background	216	9.2
	Non-EU Migration background	577	24.5
Country of origin
	Belgium (no migration background)	1,561	66.3
	Neighbouring countries	115	4.9
	Türkiye	156	6.6
	Morocco	360	15.3
	Congo	60	2.6
	Poland	49	2.1
	Mediterranean countries	52	2.2
Formal volunteering
	Formal volunteer	571	24.3
	Not a formal volunteer	1,783	75.7

Statistical Analysis

In this article, we examine social differences in participation in formal engagement among youth and how these differences either reinforce or mitigate one another. Studying interactions between different variables provides insights into which social characteristics jointly shape voluntary engagement. However, using interaction terms can quickly become complex and difficult to interpret, particularly when it requires including not only two-way interactions but also three-way and higher-order interactions (Gelman & Hill, 2006). Therefore, in a first step of the analysis, we conduct a newer methodological approach called (I-) MAIHDA, or (intersectional) multilevel analysis of individual heterogeneity and discriminatory accuracy. This approach was developed to address the limitations of specifying interaction terms, providing a more efficient and interpretable framework for intersectional quantitative research (Evans, 2015; Evans et al., 2024). While this new approach addresses some limitations of interaction models, criticisms of the MAIHDA approach persist. First, researchers state that the possible built-in collinearity between the first and second levels, and the lack of a clear indication of the negative or positive nature of the intersection effect, make interaction terms still a justified method for studying intersectionality (Lizotte et al., 2020). Second, the MAIDHA model is based on defining all possible combinations of the characteristics (which are considered as a separate level). This can only be done when we use the three-category operationalization for young people with a migration background as a more refined operationalization would lead to the problem of too many sparse cells (Evans et al., 2024). These two elements led us in a second step to complement the MAIDHA analysis with a logistic multilevel regression model with specific interaction terms.

The MAIHDA approach uses multilevel modelling to examine inequality across multiple strata, which comprise multiple crossed categories of identities or contexts (Evans et al., 2024). In this research, we will use the MAIHDA approach to examine the effects of the intersections of gender, migration background and social status on young people’s formal volunteering. MAIHDA models aim not to capture interactions but to capture the effect that occupying a specific social position has on an individual by specifying a multilevel model in which individuals are placed at the first level, and the social stratum to which they belong is placed at the second level (Evans, 2015; Evans et al., 2024).

In our study, individuals are already nested in schools. We therefore employ a cross-classified multilevel model in which individuals are simultaneously nested within two independent (i.e., non-nested) second-level clusters: schools and social strata. Because schools and social strata are not nested within one another, each individual belongs to a unique combination of both factors (Snijders & Bosker, 2012). To account for the effects of both school membership and social stratum on individual outcomes, the combined contribution of schools and strata must be treated as the total variation (Hox et al., 2018). Therefore, a cross-classified logistic MAIHDA model was specified with formal volunteering as the binary outcome variable, and gender, social status and migration background as predictors. Variance was estimated at two levels, namely schools and social strata.

In a first set of analyses (Table 2), which we will call the MAIHDA model, a null model was estimated to assess the baseline variation between schools and strata. In Model 1, the key predictors—gender, social status and migration background—were added. Because the number of respondents is too low for our sample to be divided by the country of origin, we chose to create 24 strata based on young people’s gender (two categories), social status (four categories) and migration background (three categories). This adjusted model allows us to examine the contribution of these predictors and to evaluate how much of the variance between social strata is explained by their inclusion.

OPEN IN VIEWER

Table 2.

Cross-classified Intersectional Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (n = 2,354).

		Model 0			Model 1
		OR	(95% CI)*		OR	(95% CI)*
Fixed effect: regression coefficients
Intercept
		0.25	0.19	0.33	0.37	0.23	0.58
Gender (ref: man)
	Woman				1.04	0.70	1.50
Social status (ref.: high social status
	Mid-high social status				0.95	0.60	1.58
	Mid-low social status				0.78	0.47	1.32
	Low social status				0.96	0.56	1.66
Migration background (ref.: no migration background)
	EU migration background				0.75	0.47	1.23
	Non-EU migration background				0.46	0.30	0.72
Random effects: variances
School-level		0.45	0.30	0.62	0.45	0.31	0.64
	School variance	0.20			0.21
Stratum-level		0.44	0.26	0.68	0.32	0.09	0.60
	Stratum variance	0.19			0.10
Summary statistics
VPC school		5.49%			5.74%
VPC stratum		5.25%			2.79%
Total variance partition coefficient (VPC)		10.75%			8.53%
Proportional change in variance (PCV)					48.10%
Area under receiver operating characteristic Curve (AUC)		0.67			0.67

Note: *Odds ratio is significant when 1 is not inside the 95% credible interval.

Since we found that looking at the migration background in broad categories, such as distinguishing only between people with European and non-European migration backgrounds, is reductionist, we chose to further deepen our analysis by also examining the effects of different countries of origin. In the second set of analyses (Table 3), called the interaction model, we further split the migration background into a categorical variable with seven categories based on the country of origin. We conduct a multilevel logistic regression analysis with interaction effects for the different predictors: gender (two categories), social status (four categories) and migration background (seven categories). In each model, we added an interaction term between two of the three core characteristics: gender, migration background/ethnicity and social status. We did not include a model with three-way interactions because the size of our sample leads to sparse cells.