The transition to adolescence is characterized by simultaneous changes at biological, cognitive, and social levels. During this period, adolescents become more independent in making decisions about their lives; they often withdraw from parental control and family bonding while increasing the quantity of time spent with their peers, facilitating social understanding of others (Steinberg and Morris, 2001). Moreover, adolescents achieve higher levels of sociocognitive and emotional competencies (e.g., emotion regulation, moral reasoning, empathy, perspective-taking) that enhance their prosociality (Silvers, 2022; Carlo et al., 2023; Spinrad and Eisenberg, 2023).

However, existing findings on mean-level change in the development of prosocial behavior highlighted heterogeneity in the development of prosocial behavior starting from middle childhood (Malti and Dys, 2018), which may depend on the time lag, the type of informant (e.g., self-report or teacher report), and sample characteristics (Eisenberg and Fabes, 1998). Considering children and adolescents, a meta-analysis of 179 studies conducted by Eisenberg and Fabes (1998) showed that adolescents were more likely to enact prosocial behavior than children. Accordingly, a longitudinal study by Carlo et al. (2024), involving 749 U.S. Mexican-origin adolescents (from ages 10 to 23) and using the self-report Prosocial Tendencies Measure–Revised (PMT-R; Carlo et al., 2003), showed a linear increase in the frequency of helping behaviors without apparent rewards (i.e., altruistic behaviors; Carlo et al., 2024).

In contrast, other studies have found a stable trend for prosocial behavior during adolescence. For example, a study by Griese et al. (2016), including 1,091 U.S. children (from ages 8 to 12), showed that prosocial behavior, assessed by teacher report of the Child Behavior Scale (Ladd and Profilet, 1996), followed a stable trend from late childhood to early adolescence. Moreover, a longitudinal study with an extended time lag spanning from late childhood to early adulthood (from ages 9 to 18), involving 284 children from the southwestern United States, found that prosocial behavior, measured by the prosocial subscale of the Children's Social Behavior Scale-Teacher Form (CSBS-T; Crick, 1996), remained stable during the transition to adolescence (Flynn et al., 2015). These findings suggest that prosocial behavior, which may be learned earlier during childhood, acts with greater stability in interindividual differences over time, even in the light of developmental changes and external circumstances (Eisenberg et al., 1999).

In addition, other studies showed a quadratic trend in the development of prosocial behavior. A cohort longitudinal study by Carlo et al. (2007), involving 200 adolescents and young adults from rural Eastern United States (from ages 13 to 18), showed a declining trend in self-reported prosocial behavior (i.e., using the Primary Prevention Awareness, Attitudes and Usage Scale; Swisher et al., 1984) from middle to late adolescence (around age 16) with a subsequent rebound until approximately 18 years old. Similarly, a cohort longitudinal study by Luengo Kanacri et al. (2013) explored the developmental trend of self-reported prosocial behavior of 573 Italians from early adolescence to young adulthood (from ages 13 to 21), and reported a decline in prosocial behavior using the Prosociality Scale (Pastorelli et al., 1997; Caprara et al., 2005) during this developmental period.

In summary, findings about how prosocial behavior develops are not consistent across studies, perhaps because of methodological differences, such as the type of informant (e.g., self-report or teacher ratings; Eisenberg and Fabes, 1998) and sampling characteristics (Eisenberg and Fabes, 1998). In regard to the type of informant, considering self-report measures of prosocial behavior, which are particularly relevant to the present study, previous studies point to a decreasing trend (Carlo et al., 2007; Luengo Kanacri et al., 2013). However, in regard to sampling characteristics, to the best of our knowledge, no previous longitudinal research examined the prosocial behavior of adolescents from the majority world; rather, what is known about the average developmental trajectory of prosocial behavior comes from samples from the minority world (Armstrong-Carter and Telzer, 2021). Consequently, contextualizing studies across different nations may increase understanding of whether and how adolescents behave toward others depending on their context. Thus, the present longitudinal cross-cultural study has the potential to partly address this gap by tapping into longitudinal data of the same self-reported measure of prosocial behavior collected simultaneously across six countries.

Echoing the bioecological model (Bronfenbrenner and Morris, 2007), which suggests the dynamic interplay between multiple environments at proximal (e.g., family, peers) and distal (e.g., economic and political systems) levels in determining individuals' functioning, research on behavioral development is beginning to reveal insights into how countries' characteristics, such as the sociocultural context, interact with the development of adolescents' adaptive and maladaptive behaviors (Callaghan and Corbit, 2023; Lansford, 2024). Research on prosocial behavior has underscored its malleable nature and the role of the cultural context in shaping its trajectory (de Guzman et al., 2014; Callaghan and Corbit, 2023). For example, at the proximal level, children's inclination toward prosocial actions is nurtured within familial bonds, where caregivers model and encourage behaviors that benefit others (Eisenberg et al., 2015; Pastorelli et al., 2021). Observing and being encouraged to engage in prosocial actions can help children become aware of the importance of their contributions to the social group and facilitate the acquisition of a social identity (Miller and Goodnow, 1995). At the distal level, socioeconomic factors may shape the enactment of other-oriented behavior. As reported by the Organization for Economic Cooperation and Development (OECD, 2011), comparing 140 countries worldwide, the higher the country's income, the higher the adolescents' reported enactment of prosocial behaviors, such as volunteering, donating, and helping others.

However, other findings highlight the importance of considering broader indicators to capture the wealth condition of the cultural context (Bornstein et al., 2012, 2023), such as the Human Development Index (HDI; United Nations Development Programme, UNDP, 2014). The HDI is a broader indicator than solely country income, representing a reasonable index for capturing the extent to which human development is promoted at the country level. It encompasses several levels of a country's socioeconomic growth along three dimensions: (1) long and healthy life (i.e., life expectancy at birth), (2) knowledge (i.e., expected years of schooling and mean years of schooling), and (3) decent standard of living (i.e., gross national income per capita). Existing research evidenced the validity of the HDI in predicting wellbeing and adaptive functioning in the population. For example, higher HDI has been associated with better child physical growth and educational achievement (Zietz and Rothenberg, 2022), greater child socioemotional competencies (Bornstein et al., 2022), higher nurturing and stimulating interactions between caregivers and their children (Bornstein et al., 2022), as well as less caregiver psychological aggression and physical violence (Lansford et al., 2022). Moreover, Bornstein et al. (2022), with a sample of children from 51 low- and middle-income countries, showed that HDI is positively associated with children's socioemotional development beyond the effect of positive cognitive and emotional caregiving, suggesting that a sociocultural context characterized by high HDI, in concert with characteristics of the family environment, may foster the development of children's adaptive socioemotional competencies.

Deepening the understanding of how HDI may influence prosocial development by tapping on data from the “Parenting across Cultures” (Lansford and Bornstein, 2011) project is particularly important since the six countries of this study considerably vary on many HDI dimensions. For instance, in 2009, the UNDP classified 188 countries worldwide based on the HDI dimensions. Countries with the highest stand in this ranking were those with better health, education, and living conditions. The countries participating in the present study ranged from a rank of 11 (United States) to 111 (Philippines) out of 188 countries with available data (UNDP, 2009). To give an idea of what this range entails, in 2009, children living in the United States were 48% higher than children in the Philippines in the mean years of education. Similarly, children living in Italy were 40% higher in expected years of education compared with Jordanian children. Regarding health, in 2010, the Philippines showed 80% more inequality in life expectancy compared with Italy (the largest difference), and Thailand showed 42% more inequality in life expectancy compared with the United States (the smallest difference), suggesting that in low-HDI societies there is a greater inequality in health compared with high-HDI societies. Taken together, these data evidenced the great variability in the socioeconomic macro context of the countries involved in the present study. Given these premises, the present study sought to increase the field's understanding of how national efforts in promoting population potentialities (i.e., HDI) are related to the mean level and the rate of change of prosocial behavior during adolescence.

Based on the aforementioned considerations, the goal of the present study was 2-fold:

(1) to longitudinally examine the average developmental trajectory of prosocial behavior during the transition to adolescence (from ages 9 to 16), using data from six countries: Colombia, Italy, Jordan, the Philippines, Thailand, and the United States. Although the scarcity of longitudinal studies on prosocial behavior conducted in the majority world (Armstrong-Carter and Telzer, 2021), previous studies with similar characteristics to our methods (i.e., self-report measure of prosocial behavior) and developmental period (i.e., from late childhood to adolescence), identified a declining trend in self-reported prosocial behavior during the transition to adolescence (Carlo et al., 2007; Luengo Kanacri et al., 2013). Thus, one may hypothesize that adolescents' levels of prosocial behavior show an average decline from late childhood to adolescence.

(2) to explore whether the total HDI (i.e., a composite score of life expectancy at birth, schooling, gross national income per capita) is related to the developmental trajectory of prosocial behavior during adolescence. Although previous cross-sectional findings highlighted that those children growing up in countries with lower national income also reported lower levels of prosocial behavior compared with their counterparts (OECD, 2011; House et al., 2013), to the best of our knowledge, no previous studies investigated the role of HDI in shaping prosocial behavior trajectories. Thus, we hypothesized that higher HDI is associated with higher initial levels (i.e., the intercept) of prosocial behavior. No prediction was made about whether HDI is related to changes in the trajectory of prosocial behavior during adolescence.

Finally, we tested whether gender and family SES also play a role in the developmental trajectory of prosocial behavior. Researchers previously proposed that girls are typically socialized to exhibit more caring and nurturing behaviors, whereas boys are more often encouraged to cultivate self-interest and concern for their own needs, resulting in a greater tendency of girls to enact prosocial behavior than boys (van der Graaff, 2023). Thus, we hypothesized that girls would show higher levels of prosocial behavior than boys. Regarding family SES, although a review by Robinson and Piff (2017) highlighted that individuals from lower SES show greater cooperative and compassionate tendencies than high SES individuals, predisposing these individuals to enact more prosocial behavior, other empirical evidence suggested an opposite association. For example, a study by Korndörfer et al. (2015) involving 4,020 U.S. adults showed that high SES (i.e., income, education, and occupation) is associated with a high frequency of self-reported day-helping, suggesting that individuals with high SES may be more incline in helping others during their daily life, comparing with low SES individuals. Thus, due to the mixed evidence, we did not formulate a specific hypothesis on the role of families' SES on prosocial behavior.

Participants were recruited from the longitudinal and cross-cultural study entitled: “Parenting Across Cultures” (e.g., Lansford et al., 2014). At Time 1 (T1), the present study sample included 915 children (50.5% males; M_age = 9.24, SD = 0.69, range 7–11) across six countries (Colombia, Italy, Jordan, the Philippines, Thailand, and the United States), who provided data over 7 years in 4 waves. Participants' characteristics at the following time points were: 884 participants averaged 10.34 years (SD = 0.69, range 8–13) at Time 2 (T2), 829 participants averaged 12.95 years (SD = 0.77, range 11–15) at Time 3 (T3), and 782 participants averaged 16.46 years (SD = 0.91, range 15–19) at Time 4 (T4).

The present study used data collected in each country from 2009 to 2016, and the procedures were the same at each site (see Lansford and Bornstein, 2011; Lansford et al., 2016; Skinner et al., 2017 for more details about the study procedure). Letters describing the study were sent home from schools. Those parents who agreed to be contacted to participate in the study returned a signed form.

The research team of each country conducted translation and back-translation procedures to ensure the linguistic, conceptual, and cultural equivalence of the measures used. The study received approval from Institutional Review Boards (IRBs) at participating universities in each country. The interview procedure was performed as follows. Parental informed consent and children's assent were obtained at each wave before the interview started. Participants chose the location (e.g., home, university) for the interviews. Parents were given the choice between oral and written completion of the interviews. In the initial years, interviewers carefully read each item of the questionnaire and recorded children's answers; in later years, children were given the choice between oral and written completion. For oral interviews, interviewers provided visual aids of rating scales. Interviews took approximately 2 h. In accordance with each country's IRBs, families received modest financial or other compensation for their participation.

From T1 to T4, participants' rates remained moderate-high across time. Specifically, the retention rate was 83.0% over 7 years (i.e., from T1 to T4) and across countries. The attrition rate was principally due to the children's unavailability to participate in later data collection or their refusal to participate in that specific wave. Analyses of variance showed that the missing participants at T4 did not significantly differ from their counterparts in prosocial behavior at T1 [F_{(1, 914)} = 2.54, p = 0.111], at T2 [F_{(1, 875)} = 0.59, p = 0.443], and at T3 [F_{(1, 824)} = 3.21, p = 0.074].

Regarding sociodemographic factors, mothers' years of education of missing participants were significantly lower than their non-missing counterparts [M_attrited = 11.72 vs. M_retained = 12.83; F_{(1, 864)} = 7.08, p < 0.05]. No differences between missing participants and non-missing counterparts were found for gender [F_{(1, 914)} = 0.03, p = 0.853], fathers' years of education [F_{(1, 767)} = 2.07, p = 0.151], and family income [F_{(1, 864)} = 1.31, p = 0.252].

At a preliminary level, we computed descriptive statistics and correlations (Pearson correlation coefficients) among all study variables within and across the four time points.

Given the longitudinal nature of our data, we performed a multiple-step analysis to (1) ensure the consistency of the prosocial behavior measure over time by using the Alignment Optimization (AO) recommended by Lai (2023) and (2) to model the change of prosocial behavior during the transition to adolescence (from 9 to 16 years old) by implementing a second-order Latent Growth Curve Model (LGCM; Bollen and Curran, 2006).

The AO approach is particularly suitable when one or more items measuring the construct of interest show non-invariant measurement properties in the metric and scalar model of the traditional measurement invariance approach (Vandenberg and Lance, 2000; Putnick and Bornstein, 2016), even though a configural model fit the data well (Marsh et al., 2018; Luong and Flake, 2023). Thus, using the AO allowed us to achieve a set of measurement parameter estimates that optimize non-invariant parameters while maintaining other parameters approximately invariant across time (Asparouhov and Muthén, 2014). Consequently, the AO allowed us to estimate growth factors adjusted for measurement biases in the prosocial behavior construct over time by assuming a minimum acceptable level of non-invariance (Marsh et al., 2018; Lai, 2023). Thus, we first implemented a configural longitudinal measurement invariance model, in which factor loadings, intercepts, and residual variances were freely estimated across time. Second, we extracted factor loadings and intercept estimates from the configural model. Third, we tested the AO. Fourth, to test the tenability of the AO model, we compared the configural model with the AO model using the following criteria: (a) < 30% of the pairwise comparisons using the effect size of the mean and covariance structure (d_MACS) are larger than 0.20, (b) < 50% of the items have at least one d_MACS ≥ 0.20, and (c) the R² near 1 as indicator of complete invariance (Nye and Drasgow, 2011; Asparouhov and Muthén, 2014; Lai, 2023).

To test our first aim, once the longitudinal measurement model was established, we tested a series of unconditional second-order Latent Growth Curve Models (LGCM; Bollen and Curran, 2006) by fixing the first factor loadings and intercepts at each wave to the parameters estimated using AO. In the LGCM, we estimated two additional latent factors: (1) the Intercept (I), which represents the predicted value of prosocial behavior at the initial time point and (2) the Slope (S), which represents the predicted rate of change of prosocial behavior over the four time-points (loadings set to 0, 1, 4, and 7). The LGCM allows estimating five parameters: (1) Intercept mean (I_M, i.e., the trajectory's mean value at the starting point); (2) Intercept variance (I_V, i.e., the inter-individual variability around the mean value at the starting point); (3) Slope mean (S_M, i.e., the average growth rate over time); (4) Slope variance (S_V, i.e., inter-individual variability around the average growth rate); and (5) the covariance between Intercept and Slope (I-S_C, i.e., the association between the starting point and the growth rate). The tested unconditional model is depicted in Figure 1.

To identify the trajectory that best captures the change of prosocial behavior, we tested the following three unconditional models: (1) a no-growth model or intercept only-model that estimates only the intercept and assumes a no-change trend over time of prosocial behavior; (2) a linear growth model that tests a model in which the prosocial behavior trajectory follows a linear increase or decrease across time; (3) quadratic growth model that tests a model in which the prosocial behavior trajectory assumes a quadratic shape with an initial increase or decrease and a subsequent rebound, decline, or stability. These unconditional models were compared using the Chi-square difference test (Δχ²) to identify the model that best fits our data. Finally, to explore our second and third aims referring to the effects of time-invariant predictors (i.e., HDI, gender, and family SES) on prosocial development, we tested a series of conditional models in which the growth factors (i.e., intercept and slope) were regressed on all time-invariant predictors one at a time.

We estimated parameters and handled missing data using the Maximum Likelihood estimator. To evaluate the goodness of fit of our models, we used the following indicators: Chi-square index (χ²), Comparative-Fit-Index (CFI) and Tucker-Lewis-Index values > 0.90 as indicators of acceptable model fit, and CFI/ TLI > 0.95 as indicators of good model fit (Brown, 2015; Kelloway, 1998; Kline, 1998), as well Root-Mean-Square-Error-of-Approximation (RMSEA) with 90% Confidence Interval (CI), and Standardized-Root-Mean-Square-Residual (SRMR) values lower than 0.08 as indicators of moderate model fit and below 0.05 as an indicator of good model fit (Brown, 2015; Iacobucci, 2010). Analyses were conducted in RStudio 2024.04.1, using lavaan (Rosseel, 2012).

As shown in Table 1, the youth's mean level of prosocial behavior slightly increases over time, suggesting a plausible age-related increase from late childhood to adolescence. Regarding correlational results, we found a moderate positive association between HDI at T1 and prosocial behavior at T1, T2, and T3, indicating that at higher levels of HDI, higher is prosocial behavior at ages 9, 10, and 13, and vice versa. Whereas the correlation between HDI at T1 and prosocial behavior at T4 was weak and negative, suggesting that higher previous levels of national life expectancy, education, and wealth are associated with a lower tendency of prosocial behavior enactment at age 16.

Following Lai (2023) procedure, we first implemented a series of longitudinal measurement invariance models (Vandenberg and Lance, 2000; Putnick and Bornstein, 2016) which had not reached the minimum level of longitudinal scalar invariance (see Supplementary material for results of the longitudinal measurement invariance). Because the configural, metric, and scalar model showed a good fit to the data, we implemented the alignment optimization that showed a good proportion of non-invariant parameters as reported by (a) < 30% (i.e., 5.40 out of 18) of the pairwise d_MACS larger than 0.20 (i.e., 3 out of 18: helping T1 vs. T4 = 0.32, T2 vs. T4 = 0.27, and T3 vs. T4 = 0.24), (b) < 50% of the items (i.e., 1.5 out of 3) had at least one d_MACS ≥ 0.20 (1 out of 3, i.e., helping behavior item), and (c) the R² = 0.99. Thus, based on this alignment optimization, which showed an acceptable number of parameters that held longitudinal measurement invariance, we continued implementing a second-order growth curve model based on the alignment optimization to estimate the rate of change of prosocial behavior during the transition to adolescence.

To explore the longitudinal developmental trend of prosocial behaviors during the transition to adolescence (from ages 9 to 16 years), we tested a series of unconditional second-order LGCM: no growth, linear, and quadratic change (see the Data Analytic Approach section for further details). As reported in Table 2, the linear model was chosen over the no-growth and the quadratic models because it was the model that best represented the data.

As depicted in Figure 2 and as indicated by the positive and significant mean of the slope (S_M = 0.04, p < 0.001), prosocial behavior showed a slight increase from 9 to 16 years of age, with inter-individual variability in growth over time, as shown by the significant variance of the slope (S_V = 0.01, p < 0.001). Moreover, results showed significant variance of the intercept (I_V = 0.67, p < 0.001), attesting to inter-individual variability at the starting point (i.e., at 9 years old). Finally, the covariation between the intercept and the slope was significant and negative (I-S_C = −0.07, p < 0.001), indicating that those who started with a higher initial level of prosocial behavior at 9 years of age showed less increase of prosocial behavior over time.

Our findings contribute to ongoing research on cultural factors and the development of prosocial behavior during adolescence and highlight the importance of strengthening studies on adolescents' positive development across diverse populations worldwide. The present study had several strengths. First, it included a large international sample from six countries, offering insights into the majority world, which is not frequently considered in prosocial literature. Second, we adopted a longitudinal design covering four time points over 7 years, spanning the transition to adolescence, which is a sensitive developmental window in which environmental norms and personal experiences are crucial in shaping attitudes and behaviors (Steinberg and Morris, 2001; Callaghan and Corbit, 2023). Third, in line with the bioecological model (Bronfenbrenner and Morris, 2007), our results suggest the importance of considering the role of distal sociocultural factors, such as the HDI, in shaping prosocial development.

Despite these strengths, some limitations should be noted. First, the present study relied on a brief self-report measure of prosocial behavior that reflects a general tendency to help, comfort, and share with others. A growing body of studies highlights the complexity and nuances of prosocial behavior (see Eisenberg and Spinrad, 2014) related to the relationship with the recipient (e.g., toward family members, friends, or strangers; Padilla-Walker et al., 2018). For example, a cross-sectional study by de Guzman et al. (2008) revealed that Filippino adolescents enact more prosocial behavior toward relatives, whereas Australian adolescents are more inclined to orient their prosocial behaviors toward non-relatives. Thus, future studies should consider toward whom the behavior is directed to offer a more complete picture of how prosocial behavior develops across countries. Second, although the self-report allowed us to capture the unique perspective of participants' prosocial behavior over the transition to adolescence, we recommend future studies to assess prosocial behavior by using multiple sources of informants (e.g., parents, teachers, peers) to gain insight into prosocial development from different perspectives (De Los Reyes et al., 2019). Third, our measure was originally developed in a minority world country (i.e., Italy), which might fail to capture the meaning of prosocial behavior in majority world countries (Armstrong-Carter and Telzer, 2021; Draper et al., 2023). Thus, future studies might deepen the examination of the developmental trajectory of prosocial behavior during adolescence by using different measures to capture other aspects related to the variability in prosocial behavior as a function of the cultural context (e.g., family and cultural values). Fourth, although the present study used a sampling procedure that recruited socioeconomically diverse families within countries, our sample was not representative of national populations. Therefore, results may not generalize to the entire population of the six involved countries.

Although research on the development of prosocial behavior during adolescence has prospered, several areas call for further advances. A notable gap is identifiable in the scarcity of longitudinal studies among countries of the majority world, which are underrepresented in prosocial literature. The findings of this multicultural longitudinal study attempt to partly close this gap by considering the role of sociocultural factors in relation to prosocial behavior trajectories across six countries, suggesting that the distal environment (i.e., related to life expectancy, education, and wealth at the national level) is related to how other-oriented behaviors develop as children move to adolescence. Thus, enlarging longitudinal evidence across diverse cultures and considering the role of the broad context may provide unique insights into the development of prosocial behavior and give more differentiated models for fine-grained interventions.