Promoting physical activity (PA) among youth is a significant community health priority (Aubert et al., Reference Aubert, Brazo-Sayavera, González, Janssen, Manyanga, Oyeyemi, Picard, Sherar, Turner and Tremblay2021). PA is essential not only for preventing numerous chronic adult diseases such as cancer, diabetes, and heart disease but also for enhancing mental health and overall quality of life (Marquez et al., Reference Marquez, Aguiñaga, Vásquez, Corroy, Erickson, Hillman, Stillman, Ballard, Bloodgood-Sheppard, Petruzzello, King and Powell2020; Physical Activity Guidelines Advisory Committee, 2018). Despite its importance, youth PA has declined over time, with opportunities for activity varying significantly depending on where youth live, learn, and play (Van Sluijs et al., Reference Van Sluijs, Ekelund, Crochemore-Silva, Guthold, Ha, Lubans, Oyeyemi, Ding and Katzmarzyk2021).
Ecological models are very useful in understanding influential factors related to healthy habits, such as physical activity (PA). This model emphasizes the environmental contexts of human behavior and social and psychological influences. All this can lead to a deeper understanding of the multiple influences on behavior and guide the development of intervention programs (Barnett et al., Reference Barnett, Lai, Veldman, Hardy, Cliff, Morgan, Zask, Lubans, Shultz, Ridgers, Rush, Brown and Okely2016).
Many theoretical models have been developed to explain people’s PA, but not many focus on the child population. Nevertheless, Welk (Reference Welk1999) proposed a model for promoting PA based on the developmental, behavioral, and psychological characteristics of children and adolescents. The first objective was to identify the social, personal, and environmental influences on PA in this population. Thus, the Youth Physical Activity Promotion Model (YPAP) emerged. The author of this model attempted to reorganize the influential factors of PA in several steps: (1) to identify the determinants/primary factors. The analysis of the different PA-related models showed that most of them included the same determinants grouped into the same general categories: personal, biological, psychological, social/cultural, and environmental; (2) to classify these selected determinants, four main groups were established: predisposing factors, reinforcing factors, enabling factors, and sociodemographic factors related to PA (see Figure 1); and (3) to design an intervention program based on the available resources (Welk, Reference Welk1999).
The youth physical activity promotion model (Welk, Reference Welk1999).
Figure 1. Long description
At the top center is a green oval labeled PHYSICAL ACTIVITY. Three arrows point downward from PHYSICAL ACTIVITY to three horizontally aligned ovals: left gray ENABLING FACTORS, center blue PREDISPOSING FACTORS, and right orange REINFORCING FACTORS. Each of these three ovals has arrows pointing to and from each other, forming a triangle of mutual influence. Below PREDISPOSING FACTORS are two horizontally aligned blue ovals: left AM I ABLE and right IS IT WORTH IT. Arrows connect PREDISPOSING FACTORS to both AM I ABLE and IS IT WORTH IT. ENABLING FACTORS connect downward to AM I ABLE, and REINFORCING FACTORS connect downward to IS IT WORTH IT. AM I ABLE and IS IT WORTH IT are connected by a horizontal arrow. At the bottom center is a yellow rectangle labeled PERSONAL DEMOGRAPHICS, with arrows pointing upward to ENABLING FACTORS, PREDISPOSING FACTORS, and REINFORCING FACTORS. Arrows also connect PERSONAL DEMOGRAPHICS upward to AM I ABLE and IS IT WORTH IT. All arrows indicate bidirectional or unidirectional influences among these constructs.
Predisposing factors increase the likelihood of being more active physically in adulthood. These factors answer the questions “Am I capable?” and “Is it worth it?” (Seabra et al., Reference Seabra, Maia, Seabra, Welk, Brustad and Fonseca2013; Welk, Reference Welk1999). Factors such as perceived competence are crucial when considering whether an individual will engage in PA. Several studies highlight the importance of designing intervention programs aimed at enhancing perceived competence, which in turn promotes adherence to PA (Vaquero-Solis et al., Reference Vaquero-Solis, Iglesias, Tapia-Serrano, Pulido and Sánchez-Miguel2020).
The enjoyment factor is also significant when addressing whether it is worthwhile to engage in PA. This factor, associated with PA, reflects the feeling of fun and pleasure as a positive affective response to the activity (Stevens et al., Reference Stevens, Baldwin, Bryan, Conner, Rhodes and Williams2020). Research by Bai et al. (Reference Bai, Allums-Featherston, Saint-Maurice, Welk and Candelaria2018) indicates that adolescents with higher PA levels express a greater perception of enjoyment of PA. Additionally, these individuals show fewer signs of sedentary behaviors.
Reinforcing factors focus on the direct support that young people receive for engaging in PA from the family and other close agents, including parents, friends, and Physical Education (PE) teachers. Psychosocial variables are extensively studied in the context of PA and are considered key predictors of adolescent engagement in PA (Cachón-Zagalaz et al., Reference Cachón-Zagalaz, Carrasco-Venturelli, Sánchez-Zafra and Zagalaz-Sánchez2023). Cho et al. (Reference Cho, Hussain and Kang2023) confirmed that adolescents who received greater social support had higher levels of PA. These authors included influences from parents, friends, family, and other close agents within the concept of social support.
In general, the support that young people receive from parents is associated with higher levels of PA, as parents are considered a key agent in youth development. This support is an important predictor of youth PA. Examples of parental support include transporting their children to training locations or encouraging, financing, and modeling PA through their own practices (Gómez et al., Reference Gómez, González, Pasten and Concha-Cisternas2022). Friends also influence adolescents’ PA levels because they spend much time together, playing a crucial role in behavioral changes (Hu et al., Reference Hu, Zhou, Crowley-McHattan and Liu2021).
Another significant environment where adolescents spend much of their time is school, where teachers strongly impact their students. Although the support from PE teachers is not as influential as parental support, it still plays a crucial role in their development and adherence to sports practice (Tilga et al., Reference Tilga, Kalajas-Tilga, Hein, Raudsepp and Koka2021). To achieve adequate levels of PA, it is essential to enhance adolescents’ intrinsic motivation and autonomy, a common task for PE teachers (Kalajas-Tilga et al., Reference Kalajas-Tilga, Koka, Hein, Tilga and Raudsepp2020).
However, research by Abarca-Sos et al. (Reference Abarca-Sos, Bois, Zaragoza, Generelo and Julian2013) found that not all social agents had a positive influence on adolescents’ perceived competence. Friends and fathers were positively highlighted, whereas mothers did not affect adolescents’ perceived competence, and PE teachers had a negative effect.
Therefore, it is necessary to continue studying the effects that different social agents have on young people’s PA and the relationships between them. Factors categorized as facilitators, such as body mass index (BMI), provide opportunities to practice PA and enable the population to be physically active. Studies suggest that young people accustomed to practicing PA are significantly less likely to suffer from obesity or overweight in the future (Crowe et al., Reference Crowe, Sampasa-Kanyinga, Saunders, Hamilton, Benchimol and Chaput2020). Therefore, BMI has been considered an enabling factor, as noted by Chen et al. (Reference Chen, Welk and Joens-Matre2014) in a specific study where BMI was categorized as a PA influencing factor.
Finally, among the demographic factors, we find the country of origin. While the country of origin could also be conceptualized as a broader contextual or cultural factor, we follow the approach of previous cross-national studies that have treated it as a demographic variable due to its ability to capture large-scale structural and social characteristics of the population (e.g., income level, education systems, urbanization). For example, when comparing samples from different locations, differences in PA levels are evident. Bann et al. (Reference Bann, Scholes, Fluharty and Shure2019) compared this variable in adolescents from 52 countries, finding significant differences. These differences could be attributed to the countries’ varying income levels and development policies. Therefore, we can assert that one of the determinants of adolescent PA is the country of origin, which may be more or less developed.
It is crucial to test theoretical models that explain adolescents’ PA and identify the reasons for the decline in PA during this stage. Currently, there is a lack of investigations related to the YPAP model. Wattanasit (Reference Wattanasit2009) conducted research with Thai adolescents, measuring the influence of parents, friends, perception of physical competence, and attraction to PA, which explained 15% of the variance in PA. The results indicated that the YPAP factors tested in this study could partially explain and predict PA in Thai adolescents. In the study by Crimi et al. (Reference Crimi, Hensley and Finn2009) involving American adolescents, the model considered three predictive variables: enjoyment of exercise and perceived competence as predisposing factors and parental support as a reinforcing factor. These factors explained 34% of the variance. Hilland et al. (Reference Hilland, Ridgers, Stratton and Fairclough2011) included predisposing, demographic, and biological factors in their study of English adolescents. Positive associations were found between BMI, perceived ability, self-esteem, and PA. Additionally, they highlighted the importance of addressing perceived competence and enjoyment to promote PA in schools. Chen et al. (Reference Chen, Sun and Dai2017) related self-efficacy and enjoyment with peer support and PA in Chinese teenagers. The authors noted that peer influence does not directly impact PA but does so through self-efficacy and enjoyment, with stronger effects on self-efficacy. Huard-Pelletier et al. (Reference Huard-Pelletier, Girard and Lemoyne2020) tested all the factors of the model in teenage hockey players. The results showed different behaviors depending on the type of activity, highlighting attitudes and environmental factors as key predictors for each type of behavior. Perceived competence was more strongly associated with recreational activities, whereas parental and coach support were linked to more competitive activities.
Few studies have focused on comparing these behaviors internationally. Silva et al. (Reference Silva, Lott, Wickrama, Mota and Welk2012) work involved samples from two different cultures (United States and Portugal). In both countries, social support was directly associated with moderate-to-vigorous PA (MVPA), enjoyment, and self-efficacy. However, whereas in the United States, self-efficacy and enjoyment significantly predicted MVPA, in Portugal, only self-efficacy did so. It should be noted that statistical comparisons between the samples were not possible due to the use of different measurement instruments. Other research has shown differences in PA activity prevalence and patterns between European and Latin American adolescents (Bann et al., Reference Bann, Scholes, Fluharty and Shure2019; Guthold et al., Reference Guthold, Stevens, Riley and Bull2020), with factors such as socioeconomic conditions, family roles, and educational policies shaping youth behaviors differently across regions.
Therefore, the main objectives of this study are (1) to examine the Youth Physical Activity Promotion Model (1999) in the adolescent population and (2) to analyze possible differences in adolescents according to Welk’s model (1999) depending on the country (Spain vs. Chile). The hypotheses of the study are (a) the data collected in adolescents will fit the Youth Physical Activity Promotion Model presented by Welk (Reference Welk1999); (b) there will be differences in adolescents depending on the country (Spain vs. Chile) according to Welk’s model (1999).
Method
Participants
A sample of 3,052 adolescents from Spain and Chile participated in the research: 734 were from Chile (M age = 14.74 ± 1.47), and 2318 were from Spain (Mage = 14.53 ± 1.37). On the one hand, there were 336 Chilean males (M age = 14.80 ± 1.47) and 398 Chilean females (M age = 14.70 ± 1.47) from five different high schools in the region of Valparaíso (i.e., three State high schools and two privately run high schools funded by the State). On the other hand, there were 1,180 Spanish males (M age = 14.53 ± 1.38) and 1,138 Spanish females (M age = 14.52 ± 1.36) from 14 different high schools in the region of Aragón (i.e., 11 State high schools and 3 privately run high schools funded by the State).
Measures
Physical Activity Levels
The International Physical Activity Questionnaire-Short Form (IPAQ-SF; Craig et al., Reference Craig, Marshall, Sjöström, Bauman, Booth, Ainsworth, Pratt, Ekelund, Yngve, Sallis and Oja2003) was employed. This instrument was developed for adults aged 18–65 from different countries and presented a reliability of .76 and a validity of .30. It was subsequently validated in Spanish adults (Viñas et al., Reference Viñas, Barba, Ngo and Majem2013) and adolescents (Aibar et al., Reference Aibar, García-González, Abarca-Sos, Murillo and Zaragoza2016). For this research, we used the Spanish version, composed of seven items (Hallal et al., Reference Hallal, Victora, Wells, Lima and Valle2004) that refer to PA intensity (moderate or vigorous), frequency (days per week), and duration (minutes per week) in the last 7 days.
Mother and Father Support
The Parental Support Scale (PSS, Trost et al., Reference Trost, Sallis, Pate, Freedson, Taylor and Dowda2003) was developed for adults, measuring the weekly frequency of parents’ involvement in their children’s PA. For this study, we used items adapted for Spanish adolescents (Chicote‐López et al., Reference Chicote‐López, Abarca‐Sos, Gallardo and García‐González2018). The scale comprises five items about adolescents’ perception of family support for their practice of PA or sports, referring both to the mother and the father. Items are rated on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). The adolescent-adapted version starts with the stem: your mother or father… “encourages you to practice PA or sports” or “practices PA or sports with you.” Mother’s support presented an internal consistency of (α = .78) and father’s of (α = .80).
Friends’ Support
The Sport Friendship Quality Scale (SFQS; Weiss & Smith, Reference Weiss and Smith1999) was validated for young people between 8 and 18 years to assess the perception of sport friendship quality. It is composed of five dimensions. In this research, we used “Self-esteem enhancement and supportiveness” (My friend has confidence in me during sports). This dimension has four items rated on a 5-point Likert scale ranging from 1 (absolutely false) to 5 (absolutely true). This research presented adequate internal consistency (α = .85).
Physical Activity Teacher’s Support
The Perceived Autonomy Support Scale for Exercise Settings (PASSES; Hagger et al., Reference Hagger, Chatzisarantis, Hein, Pihu, Soós and Karsai2007) was validated for young people with satisfactory values of internal consistency (α = .90). Moreno-Murcia et al. (Reference Moreno-Murcia, Rojas and Coll2008) subsequently validated the scale in Spanish adolescents, reporting adequate construct validity (.91) and reliability (.83). The scale comprises 12 items that measure adolescents’ perception of the Physical Education (PE) teacher’s support in PA or sports practice (e.g., “My PE teacher displays confidence in my ability to do active sports and/or vigorous exercise in my free time”). Responses are rated on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). In our study, internal consistency was (α = .94).
Perceived Competence
Perceived competence was evaluated through the Physical Self-Description Questionnaire (PSDQ; Marsh et al., Reference Marsh, Richards, Johnson, Roche and Tremayne1994), an instrument designed to measure different domains of physical self-concept in the adolescent population. It was validated subsequently in Spanish by Marco (Reference Marco1998), obtaining Cronbach alphas between .82 and .91. For the present study, only the six items related to perceived competence were used (e.g., “I’m good at sports and physical activity”). Each item is rated on a 6-point Likert scale ranging from 1 (false) to 6 (true). In our study, internal consistency was (α = .91).
Enjoyment
The Achievement Motivation in Schoolwork and Sport (Duda & Nicholls, Reference Duda and Nicholls1992) was used to measure perceived enjoyment of PA. We used an adapted Spanish version by Cervelló et al. (Reference Cervelló, Escartí and Balagué Gea1999), which presented adequate values of internal consistency (α = .74). This adapted version consisted of adding some words to each item (physical activity, free time, and supervised classes instead of only sports) and a reduced response Likert scale. Finally, a 5-item questionnaire was used (“I usually have fun doing physical activity and sports”), and adolescents responded on a 5-point Likert scale. The questionnaire obtained an adequate internal consistency of (α = .91).
Body Mass Index
BMI was measured through two questions that required participants to self-report their weight (in kilograms) and height (in meters). Self-reported measures have been shown to have positive and significant correlations with objective weight and height (Fonseca et al., Reference Fonseca, Silva, Matos, Esteves, Costa, Guerra and Gomes‐Pedro2010). To calculate BMI, the following mathematical formula was used: weight/(height)2.
Procedure
The same procedure was carried out in Spain and Chile. Firstly, we contacted the educational authorities of every high school in both countries. Secondly, an informed consent form was given to the parents/guardians and the participants. Thirdly, a researcher administered the questionnaire during school classes.
The questionnaire was adapted to the Chilean sample following the cultural translation guidelines developed by Sousa and Rojjanasrirat (Reference Sousa and Rojjanasrirat2011). The adaptation considered two aspects: firstly, a semantic adaptation, and secondly, a conceptual translation to simplify the expressions for the adolescents. The following procedure was used: firstly, two reviewers whose native language was Chilean Spanish reviewed the questionnaire independently. Secondly, reviewers and researchers identified the points on which they disagreed after the initial reviews. Thirdly, a pilot test was conducted with a small sample of adolescents to refine the final version.
In both countries, we ensured that data were collected during the school year (excluding holidays and exam periods) to maintain consistency in participants’ routines, physical education, and physical activity contexts.
Statistical Analysis
The data were analyzed with Mplus, Version 7.11. Structural equation models (SEM) and multigroup SEM models were developed for each country. All models were comprised of six latent variables, except for PA and BMI. PA was calculated following the validation protocol of the questionnaire (Craig et al., Reference Craig, Marshall, Sjöström, Bauman, Booth, Ainsworth, Pratt, Ekelund, Yngve, Sallis and Oja2003). To make meaningful comparisons between groups (Spain vs. Chile), the instruments must measure the same psychological constructs in both samples (Milfont & Fischer, Reference Milfont and Fischer2010). For this purpose, a factorial invariance analysis was carried out. Thus, first, a free model was established on which successively nested models were built (Little, Reference Little2013). The invariance routine begins by testing the free, unrestricted model, in which the indicator-to-construct pattern is the same in all groups (configurational invariance). This base model is then compared with the next level of factor invariance, including equality of factor loadings (weak factor invariance), equality of the intercepts of the corresponding indicators (strong factor invariance), and equality of the residual variances of the corresponding indicators (strict factor invariance). In the models, the residuals of the corresponding indicators were allowed to correlate across groups, and the first factor loading per latent variable was set to unity to establish the scale of latent variables, as recommended by Little et al. (Reference Little, Preacher, Selig and Card2007). “A ΔCFI value less than or equal to 0.01 indicates that the null hypothesis of invariance should not be rejected” according to Cheung and Rensvold (Reference Cheung and Rensvold2002).
One SEM model and two multigroup models were analyzed. Firstly, in the overall structural equation model (SEM), we tested an SEM including all participants to examine whether the proposed relationships in the YPAP model (Welk, Reference Welk1999) were supported in our adolescent sample. This model assessed the direct effects of reinforcing factors (social support from father, mother, peers, and PE teacher), predisposing factors (perceived competence and enjoyment), and the enabling factor (BMI) on PA. Secondly, to ensure comparability between Spain and Chile, we conducted a series of factorial invariance tests (Measurement invariance analysis) following Milfont and Fischer (Reference Milfont and Fischer2010). Starting from a configural invariance model, progressively more constrained models were estimated (weak, strong, and strict invariance), allowing us to verify whether constructs were measured equivalently across countries. Invariance was evaluated using ΔCFI ≤ .01 as the criterion (Cheung & Rensvold, Reference Cheung and Rensvold2002). Thirdly, after establishing measurement invariance, multigroup SEM models were tested to compare structural relationships across Spain and Chile. This step allowed us to determine whether the predictors of PA differed between countries. In these models, residuals of corresponding indicators were allowed to correlate across groups, and the first factor loading per latent variable was set to unity to identify the scale of the constructs (Little et al., Reference Little, Preacher, Selig and Card2007).
Although a multigroup SEM model does not demonstrate causality (Burkholder & Harlow, Reference Burkholder and Harlow2003), this approach allows for exploring and testing key questions, such as the pattern of relationships between groups. Considering the possible multivariate nonnormality of the measures, the robust maximum likelihood (MLR) estimator was selected for the model estimates (Wang & Wang, Reference Wang and Wang2019). Goodness-of-fit was tested with common fit indices. Thus, a model fit is considered adequate when the comparative fit index (CFI) and the Tucker–Lewis index (TLI) have values >.90, the root mean square error of approximation (RMSEA) is <.06, and the standardized root mean square residual (SRMR) is <.08 (Iacobucci, Reference Iacobucci2010).
Results
The descriptive statistics and correlation matrix for the model variables are presented in Table 1. It also includes the reliability of the latent variables of the Spanish and Chilean questionnaires, showing good results in both samples.
Means, standard deviations, reliability, and correlations between the variables under study
Table 1. Long description
The table contains twelve columns. The first column lists variables: Physical activity, S S father, S S mother, S S friends, A S P E teacher, Enjoyment, Competence, and B M I. The next three columns provide mean, standard deviation, and reliability alpha for each variable. The remaining columns show pairwise correlations, with significant values in bold. For example, Physical activity has a mean of 50.63, S S father 4.9, S S mother 4.86, S S friends 4.18, A S P E teacher 4.74, Enjoyment 4.51, Competence 4.36, and B M I 20.19. Standard deviations and reliability values are listed where available. Correlations range from negative to positive, with notable significant correlations such as S S father and S S mother at point six five, S S friends and A S P E teacher at point four eight, and Enjoyment and Competence at point five seven. Negative correlations with B M I are observed, such as Competence and B M I at negative point one seven. The table footnote defines S S as social support and A S P E teacher as autonomy support physical education teacher. Significant correlations are bolded.
Note: SS = social support; AS PE teacher: autonomy support physical education teacher. Significant correlations appear in bold.
Preliminary Analyses: Factor Invariance Analysis
Factorial invariance analyses are shown in Table 2. A decrease in CFI < .01 implies invariance. Hence, according to this criterion, weak and strong measurement invariance were supported by the group comparisons. We selected the most parsimonious model, both in terms of the invariance of the factorial loads and the intersections of the items between the groups. This implies that the groups’ measurements are considered on the same scale (equality of factorial loads) and that the scores of the groups’ elements have the same measurement metric and scale, allowing the comparison of the means of the groups’ underlying factors. Consequently, important comparisons can be made between groups (Burkholder & Harlow, Reference Burkholder and Harlow2003).
Factorial invariance analysis to ensure that the target items measure the same theoretical constructs (variables or latent factors) in both groups
Table 2. Long description
From the top row, the table lists factorial invariance levels: configural, weak, strong, and strict. For configural invariance, chi-square is 5138.057, degrees of freedom 938, RMSEA 0.045, SRMR 0.04, TLI 0.94, CFI 0.947, delta CFI and delta model are not applicable. For weak factorial invariance, chi-square is 5253.286, degrees of freedom 945, RMSEA 0.045, SRMR 0.04, TLI 0.94, CFI 0.944, delta CFI is minus 0.003, delta model is 2 versus 1. For strong factorial invariance, chi-square is 5352.663, degrees of freedom 947, RMSEA 0.045, SRMR 0.04, TLI 0.94, CFI 0.942, delta CFI is minus 0.005, delta model is 3 versus 1. For strict factorial invariance, chi-square is 5869.539, degrees of freedom 986, RMSEA 0.045, SRMR 0.04, TLI 0.94, CFI 0.931, delta CFI is 0.014, delta model is 4 versus 1. Column headers are factorial invariance, chi-square, degrees of freedom, RMSEA, SRMR, TLI, CFI, delta CFI, and delta model. The note defines all abbreviations: chi-square, degrees of freedom, root mean square of approximation, standardized root mean square residual, Tucker–Lewis index, comparative fit index, and variations in CFI.
Note: χ2 = Chi-square test; df = degrees of freedom; RMSEA = root mean square of approximation; SRMR = standardized root mean square residual; TLI = Tucker–Lewis index; CFI = comparative fit index; ΔCFI = variations in CFI.
Overall Model
The general model with the entire sample presented adequate fit to the data, χ2(1794) = 2062.442, CFI = .946, TLI = .940, RMSEA = .045, 90% CI [.043, .047], SRMR = .038. The following relationships of the model were significant: Father’s social support predicted PA (β = 0.10, p < .05); perceived competence predicted PA (β = 0.35, p < .001). Regarding the inter-relationships within the model: father’s social support was related to mother’s social support (r = .80, p < .001), friends’ support (r = .38, p < .001), PE teacher’s support (r = .34, p < .001), perception of enjoyment (r = .41, p < .001), and perceived competence (r = .44, p < .001). Mother’s social support was related to friends’ support (r = .36, p < .001), PE teacher’s support (r = .32, p < .001), enjoyment (r = .35, p < .001), and perceived competence (r = .36, p < .001). Friends’ support was related to PE teacher’s social support (r = .45, p < .001), enjoyment (r = .50, p < .001), and perceived competence (r = .41, p < .001). PE teacher’s support was related to enjoyment (r = .44, p < .001) and perceived competence (r = .37, p < .001). Finally, perceived competence and enjoyment were related to each other (r = .58, p < .001).
Model Comparison: MultiGroup Analysis by Country
After obtaining factorial invariance, a multigroup analysis by country was performed. Figure 2 presents all the standardized parameters examining the relationships in the proposed model. The model presented appropriate fit to the data; χ2(1794) = 2720.700, CFI = .942, TLI = .939, RMSEA = .046, 90% CI [.044, .048], SRMR = .043. The standardized parameters between countries are shown in Figure 2. Covariances, correlations, and nonsignificant relationships were omitted for presentation clarity.
Multigroup final model by country.
Note: **p < .01; *p < .05.
Figure 2. Long description
Top panel labeled CHILE. At the center is PHYSICAL ACTIVITY (green). To the left, BMI (gray) points to PHYSICAL ACTIVITY. Below, PERCEIVED COMPETENCE (blue) and ENJOYMENT (blue) both point to PHYSICAL ACTIVITY. The arrow from PERCEIVED COMPETENCE is labeled .29 super asterisk. From PHYSICAL ACTIVITY, four arrows extend rightward to SS FATHER, SS MOTHER, SS FRIENDS, and SS PE TEACHER (all orange), with the arrow to SS FATHER labeled .18 super asterisk. Bottom panel labeled SPAIN. Structure is identical: BMI, PERCEIVED COMPETENCE, and ENJOYMENT all point to PHYSICAL ACTIVITY. The arrow from PERCEIVED COMPETENCE is labeled .40 super asterisk. From PHYSICAL ACTIVITY, four arrows extend to SS FATHER, SS MOTHER, SS FRIENDS, and SS PE TEACHER, with the arrow to SS FATHER labeled .11 super asterisk. All other arrows are unlabeled. The asterisk indicates statistical significance.
Regarding the relationships in the model for Chile: father’s social support was related to mother’s social support (r = .70, p < .001), friends’ support (r = .31, p < .001), PE teacher’s support (r = .31, p < .001), enjoyment (r = .38, p < .001), and perceived competence (r = .41, p < .001). Mother’s social support was related to friends’ support (r = .26, p < .001), PE teacher’s support (r = .31, p < .001), enjoyment (r = .33, p < .001), and perceived competence (r = .35, p < .001). Friends’ support was related to PE teacher’s support (r = .55, p < .001), enjoyment (r = .52, p < .001), and perceived competence (r = .49, p < .001). PE teacher’s support was related to enjoyment (r = .46, p < .001) and perceived competence (r = .36, p < .001). Finally, perceived competence and enjoyment were related to each other (r = .57, p < .001).
The relationships for Spain were as follows: father’s social support was related to mother’s social support (r = .84, p < .001), friends’ support (r = .38, p < .001), PE teacher’s support (r = .35, p < .001), enjoyment (r = .41, p < .001), and perceived competence (r = .46, p < .001). Mother’s social support was related to friends’ support (r = .38, p < .001), PE teacher’s support (r = .32, p < .001), enjoyment (r = .34, p < .001), and perceived competence (r = .37, p < .001). Friends’ support was related to PE teacher’s support (r = .40, p < .001), enjoyment (r = .49, p < .001), and perceived competence (r = .38, p < .001). PE teacher’s support was related to enjoyment (r = .43, p < .001) and perceived competence (r = .37, p < .001). Finally, perceived competence and enjoyment were related to each other (r = .58, p < .001).
Discussion
The first hypothesis—the data collected in adolescents would fit the YPAP model proposed by Welk (Reference Welk1999)—is partially supported because all the study variables (PA, father’s and mother’s social support, friends’ social support, PE teacher’s social support, enjoyment, and perceived competence) are interrelated with each other, except for BMI, which is not significantly related to PA. The relationships between variables presented in the general model reflect the importance of the predisposing factors and reinforcing factors, although no association was found between enabling factors and PA. These relationships have been reported in other previous studies.
Concerning predisposing factors, perceived competence, and enjoyment are closely linked because the enjoyment or pleasure experienced while practicing any PA can promote individuals’ perceived competence, which is essential to encourage PA participation (Stankov et al., Reference Stankov, Olds and Cargo2012).
When analyses include additional factors of the YPAP model, such as social agents (facilitating factors), relationships between the variables emerge clearly. For instance, the study by Shen et al. (Reference Shen, Centeio, Garn, Martin, Kulik, Somers and McCaughtry2018) demonstrates that parents’ perceived social support is directly correlated to young people’s enjoyment of PA, particularly among individuals who perceive low competence.
Research conducted by Jekauc et al. (Reference Jekauc, Mnich, Niessner, Wunsch, Nigg, Krell-Roesch and Woll2019) tested an alternative model that establishes direct relationships between social factors—specifically, parents’ and friends’ support—and reinforcing factors, such as the perception of enjoyment, and participation in PA. This study, which involved two adolescent samples, accounted for approximately 40% of the variance in adolescents’ MVPA.
When examining the social agents who influence adolescents’ PA, authors like Haidar et al. (Reference Haidar, Ranjit, Archer and Hoelscher2019) make various distinctions. Their findings conclude that social support (encompassing both parental and peer influence) is positively associated with young people’s engagement in PA. Specifically, the study found that for every one-point increase in perceived parental support, adolescents were 1.14 times more likely to participate in structured PA on at least 5 days per week, and for every one-point increase in friends’ support, adolescents were 1.15 times more likely to engage in vigorous PA on at least 3 days per week (Haidar et al., Reference Haidar, Ranjit, Archer and Hoelscher2019).
Kılıç (Reference Kılıç2021) emphasizes the significance of friends’ and PE teachers’ perceived social support, particularly the role of positive nonverbal feedback as a predictor of adolescents’ attitudes toward PA. Despite the multitude of social agents who can influence adolescents’ PA, research indicates that parental influence remains the most significant, overshadowing the effects of other agents such as PE teachers (Olivares et al., Reference Olivares, Cossio-Bolaños, Gomez-Campos, Almonacid-Fierro and Garcia-Rubio2015). The cross-national differences found may be partially explained by cultural and socialization patterns related to parental roles. In Latin American contexts such as Chile, fathers are often perceived as role models in sports participation and tend to influence adolescents through modeling behaviors (Kastrati & Georgiev, Reference Kastrati and Georgiev2020). By contrast, in Spain, maternal involvement in adolescents’ educational and leisure activities has been reported as particularly strong, which may explain the predictive role of maternal support observed in our sample (Lawler et al., Reference Lawler, Heary, Shorter and Nixon2022). Furthermore, evidence suggests that boys and girls respond differently to parental influences, which could further nuance these findings (Neshteruk et al., Reference Neshteruk, Nezami, Nino-Tapias, Davison and Ward2017).
However, studies that have examined similar factors to those considered in this study as reinforcing factors (e.g., paternal and maternal social support, friends’ and PE teachers’ support) and predispositional factors (perceived competence) have reached varying conclusions. Whereas paternal support, peer support, and teacher’s support were positively associated with perceived competence, maternal support showed no significant influence in the findings (Chicote‐López et al., Reference Chicote‐López, Abarca‐Sos, Gallardo and García‐González2018).
Regarding enabling factors, BMI did not show a significant relationship in the general model. These results align with the study of Druică et al. (Reference Druică, Ianole-Călin, Sakizlian, Aducovschi, Dumitrescu and Sakizlian2021), where BMI was not a predictor of PA. The nonlinear relationship between PA and BMI could explain these results. A recent study showed that as BMI increases, the time spent doing PA also increases, but only up to a point, because at higher BMI scores, the relationship is negative and inverse (Druică et al., Reference Druică, Ianole-Călin, Sakizlian, Aducovschi, Dumitrescu and Sakizlian2021). Future research should consider exploring other enabling factors beyond BMI, such as access to recreational facilities, neighborhood safety, or physical fitness (Belton et al., Reference Belton, O’Brien, Meegan, Woods and Issartel2014; Welk, Reference Welk1999) to provide a more comprehensive understanding of the enabling determinants of adolescent PA.
The second hypothesis of this study posited that “There will be differences in adolescents depending on the country (Spain vs. Chile) according to the model proposed by Welk (Reference Welk1999).” The previously presented data and Figure 2 of this study show that the Chilean population had the highest values in the variables of perceived competence and paternal social support, which predict PA participation. Conversely, perceived competence and maternal social support obtained the most significant values in the Spanish population. Therefore, the second hypothesis confirms notable differences between the populations of Spain and Chile according to Welk’s (Reference Welk1999) YPAP model.
Perceived competence has proven to be a critical predictor in both countries, aligning with findings from studies such as those of Peers et al. (Reference Peers, Issartel, Behan, O’Connor and Belton2020) and Fernández-Río et al. (Reference Fernández-Río, Cecchini, Mendez-Giménez and Mendez-Alonso2018), illustrating a significant relationship between perceived competence and adolescents’ PA levels. Hamari et al. (Reference Hamari, Heinonen, Aromaa, Asanti, Koivusilta, Koski, Laaksonen, Matomäki, Pahkala, Pakarinen, Suominen and Salanterä2017) further analyzed this relationship among youths aged 10–15, finding a significant and positive correlation across all ages, with a higher significance observed in adolescents (> .001) than in younger children (> .05). Thus, perceived competence is considered a fundamental component for predicting behaviors related to adolescents’ PA participation.
The results of the current investigation also identify social agents as key factors in predicting adolescent PA. However, the specific agents who predict PA differ by population. For the Chilean sample, paternal influence is more pronounced in promoting PA among youths. This finding is consistent with a recent study conducted in Denmark, which highlighted the father’s role as a sports model as the most significant factor impacting youth PA, as opposed to the mother’s sports role (Kastrati & Georgiev, Reference Kastrati and Georgiev2020).
Nevertheless, this evidence contrasts with previous research, including a review conducted between 2009 and 2015, which found no discernible differences in fathers’ versus mothers’ influence on their children’s PA levels (Neshteruk et al., Reference Neshteruk, Nezami, Nino-Tapias, Davison and Ward2017). Given this variability of findings, future studies may benefit from analyzing parental support separately, as recent literature suggests that the influences of maternal and parental support on youth behaviors are not equivalent (Lawler et al., Reference Lawler, Heary, Shorter and Nixon2022).
Moreover, the finding that maternal and paternal support play different roles in youth physical activity has important implications for designing effective public health interventions. For instance, maternal support—often linked to emotional encouragement and organizational involvement—may be especially influential in motivating younger adolescents or girls (Gustafson & Rhodes, Reference Gustafson and Rhodes2006). In contrast, paternal support may be more closely associated with modeling behaviors and direct participation in physical activities (Edwardson & Gorely, Reference Edwardson and Gorely2010). These nuances suggest that intervention programs should be tailored not only to the adolescent’s gender and age but also to the cultural context and family structure. In contexts such as Spain and Chile, where family involvement in child development is culturally significant, engaging both mothers and fathers in youth PA promotion strategies could enhance their effectiveness (Yao & Rhodes, Reference Yao and Rhodes2015).
Limitations
Firstly, we note that all the variables analyzed were measured subjectively, as participants’ self-reports through a questionnaire. As highlighted by Hills et al. (Reference Hills, Mokhtar and Byrne2014), accelerometry provides a more precise method for measuring PA levels. Furthermore, subjective assessments of PA often overestimate actual engagement levels (Hagstromer et al., Reference Hagstromer, Ainsworth, Oja and Sjostrom2010). Regarding the calculation of BMI, while subjective reports of height and weight show a strong correlation with objective measures, the potential for participants to inaccurately report these figures remains (Brener et al., Reference Brener, McManus, Galuska, Lowry and Wechsler2003). Therefore, future studies should consider incorporating objective measures such as accelerometry to capture real-time and continuous PA data. These tools can help validate self-reported data and provide a more robust understanding of adolescent PA patterns. Nevertheless, the measurement techniques employed in this research have received scientific validation and are widely utilized.
Secondly, data collection did not occur during the same timeframe, resulting in several months of difference between various centers. This discrepancy may have influenced the samples’ PA levels. Although the IPAQ-SF questionnaire asks about an entire week, it is a school week in which adolescents have PE and extracurricular activities, without modifications throughout the school year, regardless of the season.
Thirdly, differences in sample sizes between countries (Spain, n = 2318; Chile, n = 734) should be mentioned. While both samples were sufficiently powered for statistical modeling, Meade and Bauer (Reference Meade and Bauer2007) have shown that unequal but sufficiently large samples can still yield valid and reliable results in structural equation modeling. Future research could benefit from more balanced designs to improve comparability across cultural contexts.
Fourthly, although our study was not designed to test multigroup models by sex, we recognize this as a limitation and suggest that future research examine the interplay between parental influences and adolescent gender to provide a more nuanced understanding of these relationships.
Fifthly, socioeconomic status (SES) was not collected in this study, which may act as a confounding variable in cross-country comparisons, as SES is known to influence PA behaviors in adolescents.
Lastly, it would be beneficial to include samples from a broader range of locations to enhance the robustness of the comparisons between countries, allowing for a comprehensive analysis of the cultural differences associated with living in diverse countries.
In conclusion, it can be asserted that the YPAP model proposed by Welk (Reference Welk1999) is an appropriate framework for a multifactorial analysis of influences on adolescents’ PA between the ages of 12 and 17. It is crucial to evaluate all the social factors independently, as their impacts may differ according to the specific population studied. Additionally, the context of the country emerges as a significant factor, revealing distinct differences in how various influences affect adolescents’ engagement in PA. Analyzing these factors will enable the design and implementation of multicomponent intervention programs tailored to this demography, ultimately enhancing the PA behaviors of young individuals (Messing et al., Reference Messing, Rütten, Abu-Omar, Ungerer-Röhrich, Goodwin, Burlacu and Gediga2019).
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Acknowledgments
Funding for open access charge: Universidad de Zaragoza.
Author contribution
Conceptualization: L.O.G.; Data Curation: S.R-M.; Formal Analysis: L.O.G.; Funding Acquisition: A.A-S.; Investigation: A.A-S.; Methodology: A.A-S., D.E-T., L.O.G.; Resources: A.M-D.; Supervision: A.A-S., A.M-D., L.O.G.; Writing—Original Draft: S.R-M., A.A-S.; Writing—Review and Editing: D.E-T., A.M-D., L.O.G.
Funding statement
This work was supported by a research project “Evaluation of school-based programs promoting healthy behaviors among schoolchildren in Aragón: analysis of the use of a digital tool” (grant) FUAG-006-2023 financed by the Spanish Fundación Universitaria Antonio Gargallo.
Competing interests
The authors declare none.
Open access
