Academic Stress, Academic Self-efficacy, and Psychological Distress: A Moderated Mediation of Within-person Effects

Previous research has largely failed to separate the between- and within-person effects in the longitudinal associations between academic stress, academic self-efficacy, and psychological distress (symptoms of anxiety and depression). Filling this research gap, this study investigated if academic self-efficacy mediated the relationship between academic stress and psychological distress at the intraindividual level during 3 years of upper secondary school. Gender moderation was also examined in the hypothesised model. The present sample consisted of 1508 Norwegian adolescents (baseline M age = 16.42; 52.9% high perceived family wealth; 70.6% Norwegian-born). The random intercept cross-lagged panel model results indicated (1) positive and time-invariant direct effects from academic stress to psychological distress, (2) academic self-efficacy partially mediated these effects, and (3) psychological distress impacted later academic stress. Academic stress was more strongly related to academic self-efficacy and psychological distress at the interpersonal level for boys, while the intraindividual impact of academic stress on psychological distress was stronger for girls. The study findings might have implications for school-based implementation strategies and theoretical development.


Introduction
School-related stress affects young people's quality of life (Berdida & Grande, 2022). Studies show that academic stress (Högberg et al., 2020), including demands and pressure from school (Wiklund et al., 2012) and school-related worry (Sweeting et al., 2010), impacts psychological distress (i.e. symptoms of anxiety and depression: Drapeau et al., 2012;Mirowsky & Ross, 2002) (Torsheim & Wold, 2001) over and beyond previous depressive symptoms (Murberg & Bru, 2005) on an interpersonal, between-person level. However, the intraindividual (i.e. within-person) relationship between academic stress and psychological distress, including relevant explanatory mechanisms and moderators, has largely been ignored. This study employs a moderated random intercept cross-lagged panel model (RI-CLPM) to examine the intraindividual, longitudinal associations between academic stress, academic self-efficacy, and psychological distress in a cohort of upper secondary school students.
During late secondary school, people experience increasing academic pressure from significant adults such as parents (Deb et al., 2015) and teachers (Song et al., 2015). In addition, comparing oneself to and competing with peers intensifies during this period (Eccles et al., 2003), and a series of final examinations that decide future work and educational prospects are on the horizon. In other words, students experience many day-to-day hassles related to their education, such as different pressures and demands to perform well academically during late secondary school (Dewald et al., 2014;Pascoe et al., 2020), and stressful feelings (Leonard et al., 2015;McGraw et al., 2008;Moeller et al., 2020). How adolescents experience stress is highly individual and varies in terms of duration and intensity . Motivation, performance, and well-being can increase if stressors feel challenging due to goal relevance and manageability, resulting in positive stress (eustress: Selye, 1974) (Travis et al., 2020). However, if people lack resources to cope with the various pressures and demands, the stressors are perceived as threatening and can be Self-efficacy and the Transactional Theory of Stress and Coping Lazarus and Folkman (1984) argue that people continuously go through primary and secondary cognitive appraisals, evaluating their situations and the resources available to handle them. A primary appraisal concerns the personal implications of a situation. In late secondary school, students continuously appraise their workload, namely if their school-and homework have implications for their personal well-being. There are three types of situational implications: irrelevant, benign-positive, and stressful (Lazarus & Folkman, 1984, p. 32). A stressful appraisal concerns feelings of harm/loss, threat, or challenge. Feelings of threat and challenge are most relevant to evaluating school-and homework as an implication for personal wellbeing. Threat concerns anticipation of loss or harm, such as being unable to do school-and homework and consequently receiving poor grades. A challenge is a positive situation that could lead to personal growth, such as favourable consequences for school success. Students who evaluate their school-and homework as challenging likely experience eagerness, excitement, and exhilaration. On the other hand, students who consider their school workload threatening focus on the potential harms of the situation and characteristically experience negative emotions (Lazarus & Folkman, 1984).
When students perceive their school-and homework as stressful, they must do something to cope (Lazarus & Folkman, 1984). In this case, the second appraisal becomes salient and intricately interacts with the primary appraisal to shape individuals' emotional reactions (Lazarus & Folkman, 1984). The second appraisal is an evaluation of whether the individual can manage the stressful situation. In other words, what biological, social, and cognitive resources are available to meet and cope with the contextual demands? An example of this evaluation is context-specific self-efficacy (Lazarus, 2006). Perceiving a situation as a threat might negatively inform selfefficacy through the affective/physiological state experienced in the specific setting (Bandura, 1997). The stressful reaction to school-and homework might decrease self-efficacy in the same context (i.e. academic self-efficacy), resulting in increased psychological distress (Bandura, 1997). In contrast, if the academic workload is perceived as challenging, academic self-efficacy might increase due to the positive feelings associated with school-and homework, thus reducing psychological distress.
The transactional relationship between stress, coping, and emotions is a complex system, assumed to be recursive (Lazarus & Folkman, 1987). In other words, a precursor might become an outcome and vice versa as time progresses. Therefore, in addition to the assumed associations described above, it could be beneficial to investigate the possible recursive effects over time. Specifically, psychological distress might simultaneously be an outcome and an antecedent of academic stress and academic self-efficacy. Similarly, academic self-efficacy may be an outcome and precursor of academic stress. For example, psychological distress increases stress in general (Bandura, 1997;Hammen, 2005Hammen, , 2020 and reduces academic self-efficacy (Bandura, 1997;Grøtan et al., 2019;Usher & Pajares, 2008), which can result in heightened academic stress (Chee et al., 2019;Chemers et al., 2001).

Moderating effects
According to Lazarus and Folkman (1984, p. 22), people in different groups have varying degrees of vulnerability and sensitivity to stressors and their understanding and response to them. In support of this assumption in an academic setting, Ye et al. (2018) found that gender moderated the association between academic stress and later academic self-efficacy. Specifically, they found that the association between academic stress and later academic self-efficacy was more salient for girls than boys. Moreover, studies imply that the relationship between academic stress and psychological distress is stronger for girls than boys in secondary school students (Liu & Lu, 2012;Moksnes, Moljord, et al., 2010). There is a lack of studies on the possible gender moderation of the relationship between academic self-efficacy and psychological distress. However, many studies have found gender differences in academic self-efficacy, wherein boys generally report higher levels than girls (for an overview, see Huang, 2013). These findings might imply the existence of gender differences in the association between academic self-efficacy and other factors.

Current Study
There is a lack of research on the longitudinal relationship between academic stress and psychological distress within adolescents. This study investigates the association between academic stress, academic self-efficacy, and psychological distress on an inter-and intrapersonal level throughout upper secondary school. Additionally, gender differences in these relationships are investigated. The following hypotheses are based on previous research and the assumptions of the transactional theory of stress and coping. First, academic selfefficacy will be negatively related to academic stress and psychological distress, and academic stress and psychological distress will be positively related on an interpersonal level (hypothesis 1). Second, fluctuations in academic stress will predict similar fluctuations in concurrent psychological distress (hypothesis 2). Third, fluctuations in academic self-efficacy will partially explain the association between the fluctuations in academic stress and psychological distress (hypothesis 3). Fourth, the associations between academic stress, academic self-efficacy, and psychological distress will be more salient for girls than boys (hypothesis 4). Fifth, fluctuations in psychological distress will predict opposite fluctuations in subsequent academic self-efficacy (hypothesis 5). Lastly, fluctuations in academic self-efficacy will predict an opposite fluctuation in subsequent academic stress (hypothesis 6). Due to a lack of previous research on the effect of psychological distress on later academic stress, there is no specific hypothesis regarding this relationship. However, the association is investigated in the model.

Procedure and Participants
The participants were part of the COMPLETE project (Larsen et al., 2018), a randomised controlled trial aiming to improve the psychosocial learning environment and reduce dropout rates in upper secondary school in Norway. Sixteen schools in four municipalities agreed to participate in the study. The project randomly assigned schools to one of two intervention conditions or the control group. All students who started in August 2016 in the mentioned schools were invited to participate. The participants in this study attended a general education programme, which spans 3 years of upper secondary school from grade 11 through grade 13. The study followed a cohort of students from the beginning to the end of this education. Participants (N = 1508) were adolescents who had recently started in grade 11. The respondents completed surveys in August 2016 (start of grade 11), March 2017 (end of grade 11), March 2018 (end of grade 12), and March 2019 (end of grade 13). Students who were part of the same cohort, but were absent at a previous data collection, were allowed to participate in the following data collections throughout the study. Please see Table 3 for more details on the number of participants across time points.
The Norwegian Centre for Research Data (NSD) approved that the COMPLETE project is in line with GDPR. Students under age 16 needed parental/guardian consent before participating, and respondents actively consented to participate. Ahead of participation, the students were informed about the study's aims. Researchers and research assistants in the project collected survey data using tablets on the school grounds. Students not physically present during data collection were invited to participate via e-mail.

Academic stress
The student's academic stress was measured using a single indicator from the study 'Health Behaviour in School-Aged Children (HBSC)' (Klinger et al., 2015;WHO, 2012). Participants answered how stressed they felt due to the schoolwork they must do (both work during school hours and homework). The response scale ranged from 1 (not at all) to 4 (a lot).

Academic self-efficacy
The participants' academic self-efficacy was assessed using the five-item academic efficacy scale from Patterns of Adaptive Learning Scales (PALS: Midgley et al., 2000). The scale is a context-specific measure of how capable individuals perceive themselves to be in performing and mastering schoolwork (i.e. classwork and homework). An item example is 'even if the work is hard, I can learn it'. The participants responded to the items on a Likert scale ranging from 1 (not at all confident) to 5 (very confident). Earlier studies have found acceptable Cronbach's alpha values (>0.78) (Midgley et al., 2000).

Psychological distress
Participants' psychological distress was measured by the Norwegian five-item short version of the Symptom Check List-90-R, based on the anxiety and depression subscales (Tambs & Moum, 1993). This measure is not a diagnostic tool for anxiety or depression disorders but a global indicator of mixed anxiety and depressive symptoms (Siqveland et al., 2016). Adolescents assessed how bothered or distressed they had been in the last 14 days on a scale ranging from 1 (not at all) to 4 (very much). Example indicators of depression and anxiety are 'feeling blue and sad' and 'feeling tense and worried', respectively. Previous research indicates acceptable Cronbach's values (>0.83) (Gjerde et al., 2011;Skrove et al., 2013;Strand et al., 2003;Tambs & Moum, 1993).

Gender
Gender was retrieved from registry data, coded as 0 (boys) and 1 (girls). Of note, participants also answered a question on gender identification (female, male, or other) in the questionnaires. However, very few respondents identified as non-cis or other-gendered (14 respondents on baseline). Thus, multigroup comparisons were not viable using all groups (cis females, cis males, non-cis females, non-cis males, and other-gendered).

Control variables
The following variables were included as time-invariant covariates in the model. Two dummy variables were created based on intervention conditions -participants were either in an intervention group (coded as 1) or not (coded as 0). The study measured socioeconomic position using a single indicator question on perceived family wealth (Iversen & Holsen, 2008), which was dummy coded as 0 (low) and 1 (high) by a median split. Regarding country of origin, Norwegian-born participants were coded as 0, and participants born outside of Norway were coded as 1.
This study investigated measurement invariance across time and gender using the effects-coding approach by Little et al. (2006), which is preferable to other methods (Breitsohl, 2019). In effects-coding, the average factor loadings across all indicators are constrained to 1.0, and the sum of the indicator intercepts is constrained to 0.0. The configural models were otherwise freely estimated. Equal factor loading constraints were applied across time and gender to establish metric (weak) invariance for the multiple indicator RI-CLPM (Hamaker, 2018). The invariance constraints were retained in further modelling. The academic self-efficacy and psychological distress scales achieved partial weak invariance, wherein at least two indicators of each scale were invariant over time and gender (Byrne et al., 1989). For space constraints, the measurement invariance results are presented in Table 2.

Primary analyses
The random intercept cross-lagged panel model (RI-CLPM) with academic stress, academic self-efficacy, and psychological distress was modelled following the approaches by Hamaker (2018) and Mulder and Hamaker (2021). First, each construct's random intercept (interindividual, trait-like components) was specified by adding regression coefficients from the intercepts to corresponding latent factors at each time point, constrained to 1.0. Second, 12 second-order latent factors (state-like components) were specified (one latent factor for each of the four time points in three constructs), with regression coefficients to corresponding first-order latent factors constrained to 1.0. Third, to ensure the random intercepts and within-person variables capture all variance, the variances of the first-order latent factors were constrained to 0.0. Lastly, socioeconomic position, gender, country of origin, and intervention conditions were added as control variables in the model, regressed on the random intercepts.
Academic stress was specified as a predictor of concurrent academic self-efficacy and psychological distress on an intraindividual level throughout the study period (see Fig. 1 for model specification), mainly because the first and second stress appraisals happen roughly simultaneously within individuals (Lazarus & Folkman, 1984). Further, the effect of academic self-efficacy on later academic stress and the impact of psychological distress on subsequent academic self-efficacy and academic stress was examined. A freely estimated model was compared to a time-invariant model (i.e. coefficients are equal over time). The time-invariant constraints were retained if the model fit did not significantly deteriorate the chi-square. If the constraints significantly deteriorated model fit, the constraints were not tenable and removed. Next, this study examined the academic selfefficacy mediation between academic stress and psychological distress using the "model indirect" syntax in Mplus.
To investigate if gender moderated the effects in the RI-CLPM, time-invariant constraints were initially investigated for both genders separately. Then, a multigroup analysis on the RI-CLPM with 1000 bootstraps using gender as a grouping variable was conducted, and the model constraint function in Mplus was used to compare estimates across groups.

Missingness
According to Little's missing completely at random (MCAR) test, the patterns of missingness in the study's variables were completely random (χ2 = 3092.302, df = 3031, p = 0.215). Full information maximum likelihood (FIML) was used to handle potential missing data at the construct level (Newman, 2014). Detailed information regarding the number of respondents across time is in Table 3.

Descriptive Statistics
The descriptive statistics of the study variables are presented in Table 1. There were significant gender differences in all variables. Girls experienced significantly higher academic stress, psychological distress, and lower academic self-efficacy at all times than boys. The gender differences in terms of effect sizes were, according to Cohen (1988), moderate to large concerning academic stress, moderate regarding psychological distress, and negligible to small concerning academic self-efficacy.
Random Intercept Cross-lagged Panel Model of Academic Stress, Academic Self-efficacy, and Psychological Distress The RI-CLPM of academic stress, academic self-efficacy, and psychological distress produced good model fit: .034], CFI = 0.954, SRMR = 0.039. The model included metric invariance constraints and socioeconomic position, country of origin, gender, and intervention conditions as timeinvariant covariates. Next, a fully time-invariant model with identical constraints on the regression coefficients over time was investigated. A chi-square difference test showed that the model fit significantly deteriorated (Δχ 2 = 40.658, Δdf = 21, p = 0.006). The autoregressive constraints were removed, and the time-invariant, cross-lagged constraint model was compared to the freely estimated model. The model fit did not significantly deteriorate: Δχ 2 = 24.326 Δdf = 15, p = 0.060. Therefore, the constraints were deemed tenable, and the partially time-invariant model produced good fit (χ 2 = 2266.112, df = 1046, p < 0.001, RMSEA [95% CI] = 0.032 [0.030, 0.034], CFI = 0.954, SRMR = 0.040). The results are presented in Fig. 2, and more details are in table 4.
In support of hypothesis one, the correlation between academic stress and psychological distress was positive and moderate in effect size at the interindividual level (i.e. the random intercepts) (r = 0.49, p < 0.001). Moreover, the interindividual association between psychological distress and academic self-efficacy was negative and moderate (r = −0.38, p < 0.001). Lastly, the correlation between academic self-efficacy and academic stress intercepts was negative and small (r = −0.28, p < 0.001). Thus, adolescents who experienced high academic stress throughout their upper secondary school education were also likely to experience high psychological distress and low academic self-efficacy during the same time. Additionally, individuals likely experienced opposite levels of psychological distress and academic self-efficacy during this period.
There was no support for hypotheses five or six. The results indicated a null effect between psychological distress and later academic self-efficacy. Similarly, academic selfefficacy did not impact later academic stress. However, the impact of psychological distress on subsequent academic stress was positive from T1 to T2 (β = 0.16, p < 0.001), T2 to T3 (β = 0.15, p < 0.001), and T3 to T4 (β = 0.19, p < 0.001). Thus, fluctuations in psychological distress were consistently associated with similar fluxes in academic stress approximately 1 year later throughout the study.

Gender moderation model
Before the moderation analysis of the RI-CLPM of academic stress, academic self-efficacy, and psychological distress, the appropriateness of the time-invariant constraints enforced in the mediation model was separately examined for boys and girls. The chi-square in the freely estimated RI-CLPMs was compared to the chi-square in the time-invariant constraint models in both genders. The chi-square difference tests were non-significant for both genders (p > 0.05), indicating that the time-invariant constraints were tenable. Thus, the following nine parameters between boys and girls were compared: three intercept correlation coefficients and six timeinvariant regression coefficients (academic stress on concurrent academic self-efficacy and psychological distress; academic self-efficacy on concurrent psychological distress; psychological distress on subsequent academic self-efficacy and academic stress; academic self-efficacy on subsequent academic stress).
The gender moderation RI-CLPM of academic stress, academic self-efficacy, and psychological distress achieved acceptable model fit: χ 2 = 3727.383, df = 2059, p < 0.001, RMSEA [95% CI] = 0.038 [0.036, 0.040], CFI = 0.933, SRMR = 0.057. The results are presented in Fig. 3 and table 5. In partial support of the fourth hypothesis, three parameters significantly differed across gender: the intercept correlation between academic stress and academic selfefficacy (r difference = 0.086, p = 0.025), the intercept correlation between psychological distress and academic stress (r difference = -0.082, p = 0.044), and the time-invariant regression coefficient from academic stress to concurrent psychological distress (B difference = 0.164, p = 0.000). Of note, the difference tests consider unstandardised estimates,  Please see table 5 for further details on model estimates.
The significance of the indirect effects of academic stress on concurrent psychological distress through academic selfefficacy disappeared in the moderation analysis. There were no apparent gender differences in these effects (see table 6 for details). However, the 95% confidence interval of the indirect effect did not include zero for both genders. Thus, the mediation effect, albeit small, might still be relevant for both genders despite the lack of a significant p value.

Sensitivity Analyses
This study investigated several competing models, such as different time lags between the constructs, and examined the impact of missingness on the selected model. The final model was chosen because (1) the theoretical assumptions of the transactional theory of stress and coping argue that the first and second appraisals occur simultaneously, and (2) the AIC and BIC values in the final model were lower than competing models. Regarding missingness, the final model was compared across three groups in our sample: participants with complete data (no missingness), participants with intermittent missing data patterns (non-dropouts), and all participants. The models produced similar patterns of results in terms of coefficients and standard errors.

Discussion
Few or none have investigated the associations between academic stress and psychological distress while separating inter-and intrapersonal effects. Consequently, there is little knowledge of possible explanatory mechanisms or moderators in the mentioned association on an intraindividual level. This study sought to fill that knowledge gap. The results implied that, during upper secondary school, the normative levels of academic stress, academic self-efficacy, and psychological distress were associated. Further, that academic stress consistently predicted psychological distress Standardised estimates are presented. The grey lines are nonsignificant. *** p < 0.001, ** p < 0.01, * p < 0.05 throughout the study and that academic self-efficacy partially mediated this relationship. Recursively, psychological distress impacted later academic stress. Lastly, the intraindividual association between academic stress and psychological distress was stronger for girls, while the interpersonal associations between academic stress, academic self-efficacy, and psychological distress were stronger for boys.

The Longitudinal Associations Between Academic Stress, Academic self-efficacy, and Psychological Distress
Aligning with the assumptions in the transactional theory of stress and coping (Lazarus & Folkman, 1984) and previous research, the association between academic stress and psychological distress was positive within adolescents during upper secondary school. Adolescents with, for them, unusually high (or low) academic stress at one time were increasingly likely to experience unusually high (or low) psychological distress simultaneously. Moreover, fluctuations in psychological distress were related to similar fluxes in academic stress on the following occasions. These findings indicate that interventions successful in decreasing levels of academic stress and psychological distress (e.g. Feiss et al., 2019) might lower levels in the other factor concurrently and over time, respectively. However, it might be beneficial for implementation research to investigate the effect of schoolbased measures on the intraindividual association between academic stress and psychological distress. For instance, are interventions designed on an interpersonal level effective in reducing unusually high academic stress or psychological distress at the intraindividual level? Such research might further important knowledge in the field.
Academic self-efficacy functioned as a mechanism, partially explaining the concurrent relationship between academic stress and psychological distress within adolescents over time. Indeed, fluctuations in academic stress were related to oppositional fluctuations in academic self-efficacy and similar fluxes in psychological distress simultaneously. This effect aligns with central assumptions on how selfefficacy changes within individuals, wherein adverse feelings in certain situations decrease self-efficacy in the same settings (Bandura, 1997). Because stress, as measured in this study, is an inherently negative affective state, the reduction in selfefficacy for the same context that induced the negative feeling has been explored in many instances (for an overview, see Usher & Pajares, 2008). However, the finding that fluctuations in academic self-efficacy partly explain changes in psychological distress during fluxes in academic stress is novel. Theoretical or conceptual models of stress and mental health problems might include this mechanism in adolescent samples. Even though the transactional theory of stress and coping (Lazarus & Folkman, 1984) and self-efficacy theory (Bandura, 1997) describe processes occurring within individuals, such as cognitive evaluations and change, and emotional responses, the frameworks have used research on the interpersonal level to postulate intraindividual psychological developments.
The impact of psychological distress on later academic stress was positive. Hence, fluctuations in psychological distress were associated with similar changes in academic stress~1 year later throughout the study period. Little research has focused on the impact psychological distress has on academic stress, mainly because academic stress is assumed to be an antecedent in the relationship between the two (e.g. Murberg & Bru, 2005;Tian et al., 2019). However, psychologically distressed individuals often behave in manners that create situations they perceive as stressful (Hammen, 2020). The findings in this study suggest that this effect might also apply to the educational setting, particularly the perception of school-and homework as stressful. In other words, due to an unexpected rise in psychological distress, students might behave in ways that increase the likelihood of experiencing the school-and homework as stressful later. It is possible that unusually psychologically distressed students postpone or avoid the academic workload or even physically withdraw from school. Such behaviour might result in perceiving school-and homework as a threat instead of challenging, positive, or irrelevant to personal well-being. Thus, an adverse loop of school-related stress and hopelessness, sadness, and worry might arise.

Gender Differences
Regarding gender differences, fluctuations in academic stress were more strongly associated with concurrent fluxes in psychological distress for girls than boys. The stronger intraindividual association for girls might be related to the academic pressure and demands girls perceive by others and themselves. For example, girls experience more pressures and expectations concerning their school performances (Gådin & Hammarström, 2000) and are more worried and affected by the beliefs and judgments of other people (Rudolph, 2002) than boys. Indeed, one report indicated that 39% of Norwegian girls, compared to 14% of boys, who experienced school-related stress "very often" also felt "very bothered" by symptoms of anxiety and depression (Eriksen et al., 2017). On the other hand, academic stress was significantly more strongly related to academic selfefficacy and psychological distress on an interindividual level throughout upper secondary school for boys than girls.

Limitations
One limitation is that the sample is not nationally representative. However, the participants have typical characteristics of Norwegian and Western cultures, and the results are likely transferable to other late secondary school samples similar in age and demographics.
Another possible limitation is the single-item measurement of academic stress. Single-item measures have uncertain reliability and might not adequately capture a complex psychological construct (Allen et al., 2022). A latent factor with several indicators might have provided more information concerning academic stress as a construct. However, the single indicator has been validated previously and functions well as a measure of academic stress (Klinger et al., 2015). Additionally, based on comparisons with negative stress items in stress scales, such as the perceived stress scale (Cohen et al., 1983) and the educational stress scale for adolescents (Sun et al., 2011), the included indicator is expected to have strong face validity. The bivariate correlations between the indicator across time points were moderate to strong in effect size, according to Cohen (1988), ranging from r = 0.36 (p < 0.001) to r = 0.55 (p < 0.001).
Any bias associated with self-report measures, such as common method bias (Doty & Glick, 1998) or under-and overreporting (Hunt et al., 2003;Sigmon et al., 2005), might be considered another limitation, as all data was self-reported in this study. Regarding underreporting, one study found that the difference between self-reported and administrative health service data on mood and anxiety disorders has decreased over time, particularly in adolescence (O'Donnell et al., 2016). This finding might indicate improved mental health literacy or a positive societal change in the perceptions of mental health, such as reduced stigma (O'Donnell et al., 2016). Concerning common method bias, a post hoc Harman's single factor test was performed on each time point to investigate if a latent factor was accountable for the variance in the study's data (Chang et al., 2010). The results showed that a single factor did not account for the majority of the variance, and several factor solutions were more appropriate for each measurement occasion.
Lastly, the mediating effect of academic self-efficacy between academic stress and psychological distress was small. Therefore, caution in interpreting this finding is advised. However, within-person effects tend to be smaller than effects that include both between-and within-person variances. Furthermore, the model controls for prior levels of the predictive variables. Thus, the mediation effect is relevant even though it is small.

Future Directions
Academic self-efficacy was only a partial mediator in the concurrent association between academic stress and psychological distress, implying it only explains parts of the relationship. Future research should include other relevant mediators between stress and psychological distress (e.g. coping mechanisms) in a school setting to further unravel these associations over time. Notably, researchers are encouraged to separate between-and within-person effects to truly parse the associations between academic stress and adolescent psychological distress. Moreover, when investigating the associations between academic stress, academic self-efficacy (or other mediators), and psychological distress, researchers should consider the effect of gender.

Conclusion
There is a research gap on explanatory mechanisms and moderators in the intraindividual relationship between academic stress and psychological distress during adolescence. This study aimed to fill this gap. Specifically, the inter-and intraindividual associations between academic stress, academic self-efficacy, and psychological distress, and possible gender differences in these relationships, were investigated in an upper secondary school cohort. The results showed that academic stress, directly and indirectly through academic self-efficacy, impacted concurrent psychological distress consistently during 3 years in mid-late adolescence. Psychological distress systematically affected later academic stress. Intraindividual effects were more salient for girls, and interindividual effects were stronger for boys. The study findings imply the existence of an exacerbating feedback loop between academic stress and psychological distress in upper secondary school, which functions differently for boys and girls and is partly explained by fluctuations in academic self-efficacy.
Informed consent Informed consent was obtained from all participants.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/. Two indicator factor loading constraints were removed for model fit  Est. estimate, SE standard error, CI confidence interval, RI random intercept, PD psychological distress, AS academic stress, ASE academic self-efficacy, IC intervention condition, CO country of origin, SEP socioeconomic position, G gender ***p < 0.001; **p < 0.01; *p < 0.05   Est. estimate, SE standard error, CI confidence interval, RI random intercept, PD psychological distress, AS academic stress, ASE academic selfefficacy, IC intervention condition, CO country of origin, SEP socioeconomic position, G gender ***p < 0.001; **p < 0.01; *p < 0.05