Growth mindset, persistence, and self-efficacy in early adolescents: Associations with depression, anxiety, and externalising behaviours

ABSTRACT

Growth mindset, persistence, and self-efficacy are important protective factors in understanding adolescent psychopathology, including depression, anxiety, and externalising behaviours. Previous studies have shown that dimensions of self-efficacy (academic, social, and emotional) have differential protective effects with mental health outcomes and these differences vary by sex. This study examines the dimensional mediation of self-efficacy from motivational mindsets on anxiety, depression, and externalising behaviours in a sample of early adolescents ages 10-11. Surveys were administered to participants to measure growth mindset and persistence on internalising and externalising symptoms. The Self-Efficacy Questionnaire for Children (SEQ-C) was used to measure domains of self-efficacy for mediation analysis. Multi-group structural equation modelling by sex indicated that structural paths were not invariant by sex. Significant direct effects were identified from persistence to externalising behaviours in boys, and significant direct effects were identified from growth mindset to depression in girls. In a sample of Tanzanian early adolescents, self-efficacy mediates the protective association between motivational mindsets on psychopathology. Higher academic self-efficacy was associated with reduced externalising problems in both boys and girls. Implications for adolescent programmes and future research are discussed.

KEYWORDS:

• Self-efficacy
• growth mindset
• depression
• anxiety
• externalising behaviours

Introduction

Self-efficacy has been defined as the perceived ability to complete a desired action, novel task, or to cope with a broad range of stressors (Bandura et al., 1999; Luszczynska et al., 2005). Self-efficacy is a protective factor that is integral to the process of cognitive appraisal of stressors or challenges and is related to mental health and psychological disorders (Bandura et al., 1999; Sandin et al., 2015; Schönfeld et al., 2016). The development of self-efficacy during adolescence is a function of reciprocal relationships between intrapersonal factors (affective, behavioural, and cognitive capacities) and environmental factors, such that self-efficacy influences individual patterns of behaviour with the environment and is influenced by the conditions of the environment.

Self-efficacy is important to understanding adolescent psychopathology, including depression, anxiety, and externalising behaviours (Bernstein et al., 1996; Birmaher et al., 1996). Self-efficacy can affect decision making, effort, persistence, and ability to achieve goals. Self-appraisal of self-efficacy is tied to four primary sources: actual performance, vicarious experiences, persuasion, and physiological reactions. During adolescence, social comparisons and knowledge of performance by peers may be particularly influential in shaping self-appraisal of self-efficacy (Schunk & Meece, 2006). Understanding the relationships between self-efficacy and psychopathology within the context of social norms, influenced by factors like age, gender, and culture, can provide opportunities to effectively promote mental wellbeing in adolescents.

Previous studies have reported that higher self-efficacy is associated with lower depressive symptoms, and low self-efficacy is related to higher anxiety, distress, and depression symptoms (Bandura et al., 1999; Comunian, 1989; Ehrenberg et al., 1991; Kashdan & Roberts, 2004; Kwasky & Groh, 2014; Luszczynska et al., 2005). Self-efficacy is also associated with positive mental wellbeing, optimism, and life-satisfaction (Azizli et al., 2015; Luszczynska et al., 2005). While the mediating role of general self-efficacy has been studied most in terms of mediating stressful life events on mental wellbeing, research is needed to understand how self-efficacy may mediate protective factors. Identifying modifiable protective factors is important for early adolescent prevention programmes that seek to reduce risk for mental health problems that emerge in mid- to late-adolescence.

While previous research has indicated the mediating role of general self-efficacy in a broad range of areas, a domain specific analysis of self-efficacy is needed to better differentiate self-efficacy domains and relationships with mental wellbeing (Andretta & McKay, 2020). Self-efficacy can be broken down into three domains: academic, social, and emotional. Academic self-efficacy refers to the individual’s perceived ability to control their learning behaviours, master subjects, and meet scholastic expectancies; social self-efficacy refers to the individual’s perceived ability to be authentic and assertive in peer relationships; and emotional self-efficacy refers to the individual’s perceived ability to cope with negative emotions (Muris, 2001).

Research indicates that self-efficacy acts as a mediator between stress and psychopathology (Maciejewski et al., 2000). More recently, research has explored how domains of self-efficacy are related to mental health. One study found that academic and emotional self-efficacy were significantly negatively correlated with depressive symptoms (Muris, 2001). Other studies have shown that academic self-efficacy was the best predictor of depression symptoms in comparison to social and emotional self-efficacy (Tahmassian & Anari, 2009). Higher emotional self-efficacy has been associated with higher levels of wellbeing and lower psychological symptoms (Andretta & McKay, 2020; Wigelsworth et al., 2017). Higher social self-efficacy has been shown to be associated with reduced symptoms of depression, and lower social and emotional self-efficacy has been shown to be associated with greater anxiety symptoms and social phobias (Ahmad et al., 2014; Muris, 2002). Analysis of domains of self-efficacy and relationships with mental health can increase the precision of intervention approaches that target protective factors for mental health (Kannangara et al., 2018).

Self-efficacy is closely related to motivational proclivities, including motivational mindsets or mental frameworks for assigning meanings to events and can enhance motivation (Burnette et al., 2020; Luszczynska et al., 2005). Two motivational mindsets that have received greater attention by researchers are growth mindset and persistence. Persistence is defined as intentional and continued action towards a particular goal and is similar to the concept of grit which combines persistence with passion for goal attainment (Duckworth & Quinn, 2009; Lufi & Cohen, 1987). Growth mindset represents a set of beliefs that intelligence or personality are human attributes that can change across the lifespan through effort, practice, and education (Dweck, 2016). In contrast, a fixed mindset is a set of beliefs that human attributes are fixed qualities that cannot be changed (Burnette et al., 2020). Growth mindset has been associated with well-being outcomes whereas a fixed mindset has been associated with helplessness and resistance to confronting challenges (Dweck, 2016; Dweck & Yeager, 2019).

Motivational mindsets are correlated with self-efficacy and can together can be important protective factors for internalising and externalising symptoms (Schleider et al., 2015). These mindsets vary across a range of attributes and abilities and may be situationally dependent (type of stressors encountered) and context dependent (environmental factors shaping perceptions of stress). Mindsets are believed to buffer against the negative impact of adverse life experiences and perceptions of self-efficacy can empower individuals to be able to cope with stressors (De Castella et al., 2014). Self-efficacy is associated with individual task choices, effort, self-regulation, and achievement (Bandura et al., 1999; Luszczynska et al., 2005; Schunk & Meece, 2006). Both self-efficacy and mindset are important factors to understanding how individuals perceive themselves, express emotions, and make behavioural choices.

The protective effects of motivational mindsets and self-efficacy on mental health are tied to individual self-appraisal of these measures and may be shaped by social norms and expectations. Previous research has found significant gender differences in self-efficacy, with girls reporting lower levels of emotional self-efficacy than boys (Muris, 2001). Previous systematic reviews have shown that girls often face more restrictions in voice and personal agency than boys, with feelings of restriction increasing with the onset of puberty (Kågesten et al., 2016). Changes in gender roles often shift during the pubertal transition, highlighting the importance of targeting protective factors during early adolescence to reduce gender disparities that amplify in mid- to late-adolescence.

While research has examined the mediating role of self-efficacy in response to risk factors, additional research is needed to study the mediating role of self-efficacy from protective factors. Representation of evidence from low-resource populations in the global south are important because in low-resource populations, risk factors may be more difficult to modify than protective factors. This study hypothesises that self-efficacy will mediate the protective effect of growth mindset and persistence on anxiety, depression, and externalising behaviours and these effects will vary by gender during early adolescence. Specifically, we hypothesise that mediation effects will reduce internalising symptoms in girls (depression and anxiety) and externalising symptoms in boys.

Methods

Study procedures and sample selection

This study utilised data collected from Discover Learning, a three-arm comparative effectiveness trial of an intervention to support early adolescents’ social emotional skills and mindsets. Participants were recruited from the peri-urban Temeke Municipality in Dar es Salaam, Tanzania. Approximately 14.9 million people in Tanzania live below the international poverty line of USD $1.90 per day (World Bank, 2019). Dar es Salaam is Tanzania’s largest city and rapidly growing with a population of about 7.4 million people as of 2022 (Central Intelligence Agency, 2022). Urban areas like Dar es Salaam have experienced the greatest decline in poverty in conjunction with improvements in living conditions and electricity between 2012 and 2018 (World Bank, 2019). However, human capital remained low in these areas with a Human Capital Index of 0.4 in 2018 (World Bank, 2019).

A detailed study protocol of the Discover parent study can be found in a separate publication, which includes additional information regarding the successive three-arm trial, participant recruitment, and eligibility conditions (Cherewick et al., 2020). Results indicate that 66.6% of participants lived with both parents, 33.4% lived with one parent and the mean household size was 5.7 (SD = 2.6) (Cherewick et al., 2021). The Tanzanian poverty scorecard was used to estimate socioeconomic status with higher scores indicating higher wealth (Schreiner, 2012). In this sample, the mean score on the Tanzanian poverty scorecard was 61.4 (SD = 11.6), with a possible range of 0–71 (Cherewick et al., 2021). Analyses were performed using a sample of 579 young adolescents (309 girls and 270 boys) ages 9–12 years (mean age = 10.48; SD = 0.55). Written consent from parents/caregivers and verbal assent from all adolescent participants were provided before the survey was administered, which took approximately 45-60 min to complete. Participants were compensated appropriately with a gift of a notebook and pencil (valued at less than USD $2) based on consultations with the research team, teachers, caregivers, and community members. The data used in this study was collected at baseline prior to the Discover intervention in June-July 2019. This study presents a cross-sectional view of associations between measures of self-efficacy, persistence, growth mindset, and internalising/externalising symptoms.

Survey measures

All participants completed a survey at baseline, prior to implementing the Discover Learning intervention. Survey questions included demographic measures (e.g. age, gender), social emotional mindsets and skills (e.g. self-efficacy, persistence, growth mindset), and psychological assessment (i.e. anxiety, depression, externalising behaviours). Measurement scales were selected based on their use in previous studies of adolescent populations in low- and middle-income countries and were adapted for the age range of the participants.

Self-efficacy questionnaire for children

Self-efficacy was measured using the Self-Efficacy Questionnaire for Children (SEQ-C) scale (Muris, 2001). Participants responded to 24 items measuring three different dimensions of self-efficacy: academic, social, and emotional. Participants rated how well each statement describes them using a 5-point scale (1 = ‘Not at all’, 2 = ‘A little bit’, 3 = ‘About average’, 4 = ‘Well’, 5 = ‘Very well’). Average scores were generated, with higher scores indicating greater levels of self-efficacy. Table 1 provides the adapted statements of the SEQ-C used in this study’s questionnaire. Internal consistency was acceptable in this study for academic (α = 0.69), social (α = 0.65), and emotional self-efficacy (α = 0.77).

Table 1. Self-efficacy questionnaire for children (SEQ-C).

Subscale	Measure item
Academic	1. I can ask teachers to help me when I get stuck on schoolwork.
	2. I can study even if there are other interesting things to do.
	3. I can study for a test.
	4. I can finish my homework every day.
	5. I can pay attention during every class.
	6. I can succeed in understanding all subjects in school.
	7. I succeed in making my parents satisfied with my schoolwork.
	8. I can succeed in passing tests at school.
Emotional	1. I can succeed at cheering myself up when something bad has happened to me.
	2. I can succeed in being calm even when I’m scared.
	3. I can stop myself from being nervous.
	4. I can control my feelings.
	5. I can give myself a pep talk when I feel low.
	6. I can tell a friend I don’t feel well.
	7. I can succeed in stopping unpleasant thoughts.
	8. I can succeed in not worrying that bad things might happen.
Social	1. I can express my opinions when others disagree with me.
	2. I can become friends with other children.
	3. I can chat with an unfamiliar person.
	4. I can work in harmony with my classmates.
	5. I can tell other children that they are doing something I don’t like.
	6. I can tell a funny event to a group of children.
	7. I can succeed in staying friends with other children.
	8. I can prevent arguments with other children.

Growth mindset

Growth mindset was assessed using a combination of items from Dweck’s Theories of Intelligence Scale (Dweck, 2013, 2016; Ingebrigtsen, 2018) and adapted growth mindset oriented items from the California Healthy Kids Survey. Items were adapted for understanding and selected based on the age and context of the study participants. Participants responded to nine items, ranking their agreement with each statement on a 4-point Likert scale of ‘Strongly Disagree’ to ‘Strongly Agree’. Items included statements such as: ‘I have a certain amount of intelligence and I really can’t do much to change it’, ‘I can learn new things’, and ‘the harder I work at something, the better I will be at it’. Average scores were generated, with higher scores indicating greater levels of growth mindset. Cronbach’s alpha was α = 0.70 in this sample (Constantine & Benard, 2001).

Persistence

Persistence was measured with the Persistence Scale for Children (Lufi & Cohen, 1987). Participants responded to 10 items, ranking their agreement with each statement on a 4-point Likert scale (1 = ‘Strongly Disagree’, 2 = ‘Disagree’, 3 = ‘Agree’, 4 = ‘Strongly Agree’). Items included statements such as ‘even if I fail to solve a problem, I try again and again and hope that I will find the solution’, ‘when I read a book, I do not skip any pages’, and ‘I like to finish all my homework on time, even though sometimes it is hard and I’d rather go play’. Average scores across items were generated, with higher scores indicating greater levels of persistence. Cronbach’s alpha indicates good internal reliability (α = 0.66) (Lufi & Cohen, 1987).

African youth psychological assessment

The African Youth Psychological Assessment (AYPA) was used to assess how participants internalise (i.e. anxiety, depression) and externalise problems, and is used to measure psychosocial health in African youth (Betancourt et al., 2014). The adapted version of this scale used in this study is described in Table 2. For each statement, participants were asked to describe how frequently they would display certain behaviours on a 4-point Likert scale (1 = ‘Never’, 2 = ‘Somewhat’, 3 = ‘Often’, 4 = ‘All the time’). Average scores were generated with higher scores indicating greater levels of psychosocial symptoms. Cronbach’s alpha indicates good internal reliability (α = 0.72–0.88) (Betancourt et al., 2014).

Table 2. African youth psychological assessment (AYPA).

Subscale	Items
Anxiety	I worry constantly, I do not sleep at night, I have a lot of worries, I think people are chasing me, I cry continuously
Depression	I have a lot of thoughts, I have pain all over my body, I think I am of no use, I think about suicide, I sit alone, I get headaches, I feel a lot of pain in my heart, I sit with my cheek in my palm, I cry when I am alone, I feel cold, I want to be alone, I hold my head, I drink alcohol, I do not think straight, I mutter to myself, I feel I can do nothing to help myself, I feel sad, I think of bad things
Externalising Behaviours	I am disobedient, I insult friends, I fight, I use bad language, I am disrespectful, I misbehave, I deceive, I am a rough person, I consume drugs (like marijuana)

Data analysis

Data were analyzed using Stata Statistical Software Version 17 (StataCorp, 2021). Sample characteristics are presented using frequencies and means by sex. Additional variables were examined as potential confounders and assessed for relationships with independent and outcome variables. Descriptive statistics compared each variable included in the structural equation models by gender to test for significant differences. Pearson’s correlations were calculated between all measured variables and tested for significance. Structural equation modelling (SEM) was used to estimate associations between growth mindset, persistence, and dimensions of self-efficacy on depression, anxiety, and externalising behaviours. Model invariance by sex was tested using multi-group structural equation modelling. Examination of SEM model fit indices were used to drive refinement of SEMs and improve model fit. Goodness of fit was evaluated using Chi-square, the root mean square error of approximation (RMSEA), the comparative fit index (CFI), the Tucker-Lewis Index (TLI) and the standardised root mean residual (SRMR) (Bentler, 1990; Steiger, 1980; Tucker & Lewis, 1973). The following statistical criteria was used to evaluate model fit: RMSEA < 0.06; CFI > 0.90; TLI > 0.90; and SRMR < 0.08 (Kline & Santor, 1999). The maximum likelihood method of estimation was used to estimate the SEM. Modification indices were examined to revise and improve the fit of the model (Kline & Santor, 1999).

Ethics approval

This study was approved by the University of California Berkeley Committee for Protection of Human Subjects Institutional Review Board (IRB) - (CPHS Protocol Number: 2018-01-10628); in June 2018. The primary local partner in Tanzania, Health for a Prosperous Nation, obtained ethical clearance for these research activities from the National Institute of Medical Research – the local IRB in Tanzania (Ref. NIMR/HQ/R.8a/Vol. IX/2851) in August 2018.

Results

Sample characteristics

Five hundred and seventy-nine (579) adolescents ages 9–12 years were included in the analytic sample of this study (Table 3). The sample included 270 boys (46.6%) and 309 girls (53.4%). The mean age of study participants was 10.48 years (SD = 0.55). 172 (29.7%) of participants were in 3rd grade, 261 (45.1%) in 4th grade, and 146 (25.2%) in 5th grade.

Table 3. Sample characteristics (N = 579).

	Girls		Boys		Total
	n	%	n	%	n	%
Age (years)
9	7	2.3	6	2.2	13	2.3
10	139	45.0	137	50.7	276	47.7
11	161	52.1	127	47.0	288	49.7
12	2	0.6	0	0	2	0.4
Mean Age (SD)	10.5 (0.56)		10.4 (0.54)		10.48 (0.55)
Grade
3	77	24.9	95	35.2	172	29.7
4	142	46.0	119	44.1	261	45.1
5	90	29.1	56	20.7	146	25.2

Correlations between key variables are described in Table 4. Several variables were statistically significantly correlated, although effect sizes were relatively small. Emotional self-efficacy was negatively correlated with female gender (r = −0.10; p < 0.05). Growth mindset was positively correlated with age (r = 0.09; p < 0.05), academic self-efficacy (r = 0.27; p < 0.001), social self-efficacy (r = 0.31; p < 0.001), and emotional self-efficacy (r = 0.23; p < 0.001). Persistence was positively correlated with female gender (r = 0.10; p < 0.05), academic self-efficacy (r = 0.42; p < 0.001), social self-efficacy (r = 0.34; p < 0.001), emotional self-efficacy (r = 0.34; p < 0.001), and growth mindset (r = 0.38; p < 0.001). Depression was significantly negatively correlated with social self-efficacy (r = −0.11; p < 0.01) and growth mindset (r = −0.17; p < 0.001). Anxiety was significantly negatively correlated with growth mindset (r = −0.12; p < 0.01). Externalising symptoms were significantly negatively correlated with academic self-efficacy (r = −0.34; p < 0.001), social self-efficacy (r = −0.19; p < 0.001), emotional self-efficacy (r = −0.12; p < 0.01), growth mindset (r = −0.16; p < 0.001), and persistence (r = −0.22; p < 0.001).

Table 4. Pearson’s correlations between key variables.

	Key variables	1	2	3	4	5	6	7	8	9	10
1	Sex	–
2	Age	0.06	–
3	Academic self-efficacy	0.05	0.00	–
4	Social self-efficacy	0.02	0.01	0.59***	–
5	Emotional self-efficacy	−0.10*	0.01	0.53***	0.61***	–
6	Growth mindset	0.07	0.09*	0.27***	0.31***	0.23***	–
7	Persistence	0.10*	0.07	0.42***	0.34***	0.34***	0.38***	–
8	Depression	0.01	0.01	−0.08	−0.11**	−0.02	−0.17***	−0.03	–
9	Anxiety	0.01	0.04	0.01	−0.01	0.03	−0.12**	0.00	0.75***	–
10	Externalising symptoms	−0.03	0.03	−0.34***	−0.19***	−0.12**	−0.16***	−0.22***	0.51***	0.43***	–

*Correlation is significant at p < 0.05 level; **Correlation is significant at p < 0.01 level; ***Correlation is significant at p < 0.001 level.

Structural equation models

The structural equation models fitted included variables based on previous research indicating significant relationships with mental health and wellbeing outcomes. Additional variables were tested but were non-significant (number of friends, days missed from school, village violence), and were excluded to achieve a parsimonious model. Table 5 displays descriptive statistics of included variables by gender. Girls had significantly lower emotional self-efficacy in comparison to boys (β = 1.22; p = 0.013) and higher scores on the persistence measure than boys (β = 0.73; p = 0.021).

Table 5. Descriptive statistics for measured variables.

		Male	Female	β	p-value	Total
		Mean (SD)	Mean (SD)			Mean (SD)
Self-efficacy
	Academic self-efficacy	33.4 (0.24)	33.8 (0.22)	0.38	0.245	33.6 (3.90)
	Social self-efficacy	30.2 (0.26)	30.4 (0.27)	0.16	0.678	30.3 (4.52)
	Emotional self-efficacy	30.1 (0.96)	28.9 (0.36)	1.22	0.013*	29.5 (5.93)
Mindsets
Growth mindset		18.1 (0.18)	18.6 (0.19)	0.46	0.085	18.4 (0.13)
Persistence		32.0 (0.22)	32.8 (0.23)	0.73	0.021*	32.4 (0.16)
Psychological symptoms
	Depression	4.1 (0.30)	4.2 (0.22)	0.10	0.818	4.16 (0.22)
	Anxiety	1.2 (0.12)	1.2 (0.11)	0.03	0.854	1.22 (0.08)
	Externalising behaviours	1.02 (0.14)	0.9 (0.12)	0.12	0.505	0.96 (0.09)

p-values in bold are significant; * < 0.05; ** < 0.01; *** < 0.001.

Initial testing of the SEM examined used the method of maximum likelihood estimation. To obtain the best model, six alternative models were fitted (Table 6). First, the hypothesised mediation model was fitted with growth mindset and persistence mediated by academic, social, and emotional self-efficacy on depression, anxiety, and externalising behaviours (Model 1). Results of Model 1 indicated model fit was marginally acceptable (χ²= 1498; df = 27; p < 0.001; CFI = 0.987; TLI = 0.943; RMSEA = 0.073 [90% CI: 0.045, 0.104]; SRMR = 0.033). Next, Model 2 examined invariance by sex using multi-group structural equation modelling that constrained path coefficients to be equal between boys and girls. Model fit indices indicated the SEM was not invariant by sex (χ ²= 1559; df = 54; p < 0.001; CFI = 0.958; TLI = 0.949; RMSEA = 0.070 [90% CI: 0.053, 0.087]; SRMR = 0.107). For this reason, subsequent SEM models were fitted separately for girls and boys (Models 3-6). The SEM model fitted for boys (Model 3) was inadequate for the TLI and RMSEA (χ²= 671; df = 27; p < 0.001; CFI = 0.981; TLI = 0.914; RMSEA = 0.087 [90% CI: 0.044, 0.134]; SRMR = 0.039). Examination of modification indices indicated that model fit would be improved by including a direct path from persistence to externalising behaviours. The revised SEM for boys with this direct path (Model 4) resulted in excellent model fit (χ²= 671; df = 27; p < 0.001; CFI = 0.997; TLI = 0.986; RMSEA = 0.035 [90% CI: < 0.001, 0.097]; SRMR = 0.028). The SEM fitted for girls (Model 5) was acceptable (χ² = 888; df = 27; p < 0.001; CFI = 0.993; TLI = 0.971; RMSEA = 0.055 [90% CI: < 0.001, 0.102]; SRMR = 0.031). However, examination of modification indices indicated that model fit would be improved with the inclusion of a direct path from growth mindset to depression. The revised model including this direct path (Model 6) improved model fit and resulted in excellent fit as reflected by model fit indices (χ² = 888; df = 27; p < 0.001; CFI = 0.998; TLI = 0.988; RMSEA = 0.036 [90% CI: < 0.001, 0.093]; SRMR = 0.024).

Table 6. Multigroup SEM model invariance and SEM models by sex.

Model	Multigroup SEM	RMSEA 90% CI								Δ From previous model
		χ^2 (df)	CFI	TLI	RMSEA	LB	UB	SRMR	p-value	Δ χ^2	Δ CFI	Δ RMSEA
1	Unconstrained	1498 (27)	0.987	0.943	0.073	0.045	0.104	0.033	<0.001	–	–	–
2	Constrained	1559 (54)	0.958	0.949	0.070	0.053	0.087	0.107	<0.001	62 (27)	−0.029	−0.003
	SEM models by sex
3	SEM boys	671 (27)	0.981	0.914	0.087	0.044	0.134	0.039	<0.001	–	–	–
4	SEM boys revised^a	671 (27)	0.997	0.986	0.035	<0.001	0.097	0.028	<0.001	0 (0)	0.016	−0.052
5	SEM girls	888 (27)	0.993	0.971	0.055	<0.001	0.102	0.031	<0.001	–	–	–
6	SEM girls revised^b	888 (27)	0.998	0.988	0.036	<0.001	0.093	0.024	<0.001	0(0)	0.005	−0.019

Note: SEM = structural equation model, LB = lower bound, UB = upper bound.

^a Revised model includes direct path from persistence to externalising behaviours.

^b Revised model includes direct path from growth mindset to depression.

Figure 1 presents the SEM results for girls. Table 7 presents the standardised path coefficients, p-values, and 95% confidence intervals for the structural paths between latent variables of persistence and growth mindset; dimensions of self-efficacy; and depression, anxiety, and externalising behaviours. Academic self-efficacy was positively associated with both growth mindset (β = 0.13; p = 0.021) and persistence (β = 0.41; p < 0.001). Social self-efficacy was associated with growth mindset (β = 0.21; p < 0.001) and persistence (β = 0.37; p < 0.001). Emotional self-efficacy was not associated with growth mindset; however, emotional self-efficacy was positively associated with persistence (β = 0.39; p < 0.001). Academic self-efficacy was found to be associated with lower externalising symptoms among girls (β = −0.41; p < 0.001). Although not statistically significant, modification indices supported inclusion of a negative path of association between growth mindset and depression among girls (β = −0.08; p = 0.079).

Figure 1. Structural equation model for girls. *CFI = 0.998; TLI = 0.988; SRMR = 0.024; RMSEA = 0.036.

Table 7. Standardised path coefficients between mindsets, self-efficacy and mental health in girls.

Structural paths	Coefficient	SE^1	Z	p-value	95% Confidence Interval
Self-efficacy
Academic self-efficacy
Growth mindset	0.13	0.06	2.30	0.021*	0.02	0.24
Persistence	0.41	0.06	7.16	<0.001***	0.29	0.52
Social self-efficacy
Growth mindset	0.21	0.05	4.00	<0.001***	0.11	0.31
Persistence	0.37	0.05	6.90	<0.001***	0.26	0.47
Emotional self-efficacy
Growth mindset	0.10	0.06	1.82	0.068	−0.01	0.22
Persistence	0.39	0.05	7.30	<0.001***	0.28	0.49
Mental health
Depression
Academic self-efficacy	0.02	0.08	0.29	0.772	−0.14	0.19
Social self-efficacy	−0.15	0.09	−1.75	0.080	−0.32	0.02
Emotional self-efficacy	0.06	0.07	0.93	0.354	−0.07	0.20
Growth mindset	−0.08	0.05	−1.76	0.079	−0.18	0.01
Anxiety
Academic self-efficacy	0.00	0.09	0.05	0.961	−0.17	0.18
Social self-efficacy	0.00	0.09	−0.03	0.979	−0.18	0.17
Emotional self-efficacy	0.01	0.07	0.10	0.920	−0.13	0.14
Externalising behaviours
Academic self-efficacy	−0.41	0.10	−4.05	<0.001***	−0.61	−0.21
Social self-efficacy	−0.01	0.09	−0.17	0.867	−0.18	0.15
Emotional self-efficacy	0.10	0.07	1.58	0.113	−0.02	0.23
Covariance
(Academic self-efficacy, social self-efficacy)	0.56	0.05	12.39	<0.001***	0.47	0.65
(Academic self-efficacy, emotional self-efficacy)	0.41	0.05	8.00	<0.001***	0.31	0.51
(Social self-efficacy, emotional self-efficacy)	0.58	0.04	13.11	<0.001***	0.49	0.66
(Depression, anxiety)	0.72	0.04	18.00	<0.001***	0.65	0.80
(Depression, externalising symptoms)	0.57	0.06	9.07	<0.001***	0.45	0.69
(Anxiety, externalising symptoms)	0.47	0.05	8.72	<0.001***	0.36	0.57

Note. *p < 0.05 **p < 0.01 ***p < 0.001; All measurement model paths were significant at the p < 0.001 level; ¹Robust Standard Errors reported.

Figure 2 presents the SEM results for boys. Table 8 presents the standardised path coefficients, p-values, and 95% confidence intervals for the structural paths between latent variables of protective factors; dimensions of self-efficacy; and depression, anxiety, and externalising behaviours among boys. Academic self-efficacy was associated with both growth mindset (β = 0.13; p = 0.028) and persistence (β = 0.32; p < 0.001). Social self-efficacy was associated with both growth mindset (β = 0.22; p < 0.001) and persistence (β = 0.13; p = 0.030). Unlike in the SEM model with girls, for boys, emotional self-efficacy was positively associated with growth mindset (β = 0.15; p = 0.016) and persistence (β = 0.19; p = 0.003). Academic self-efficacy was found to be associated with lower externalising symptoms among boys (β = −0.26; p < 0.001). The direct path between persistence and externalising symptoms was also significant (β = −0.26; p < 0.001).

Figure 2. Structural equation model for boys. *CFI = 0.997; TLI = 0.986; SRMR = 0.028; RMSEA = 0.035.

Table 8. Standardised path coefficients between mindsets, self-efficacy and mental health in boys.

Structural paths	Coefficient	SE^1	Z	p >|z|	95% Confidence interval
Academic self efficacy
Growth mindset	0.13	0.06	2.19	0.028*	0.01	0.25
Persistence	0.32	0.06	5.46	<0.001***	0.21	0.44
Social self efficacy
Growth mindset	0.22	0.05	3.96	<0.001***	0.11	0.33
Persistence	0.13	0.06	2.17	0.030*	0.01	0.24
Emotional self efficacy
Growth mindset	0.15	0.06	2.40	0.016*	0.03	0.26
Persistence	0.19	0.06	2.99	0.003**	0.06	0.31
Mental health
Depression
Academic self efficacy	−0.11	0.08	−1.30	0.194	−0.27	0.06
Social self efficacy	−0.11	0.09	−1.25	0.212	−0.27	0.06
Emotional self efficacy	0.13	0.07	1.77	0.078	−0.01	0.27
Anxiety
Academic self efficacy	0.00	0.09	−0.02	0.987	−0.18	0.17
Social self efficacy	−0.09	0.09	−1.03	0.302	−0.27	0.06
Emotional self efficacy	0.12	0.07	1.58	0.115	−0.03	0.26
Externalising Behaviours
Academic self efficacy	−0.26	0.07	−3.69	<0.001***	−0.40	0.12
Social self efficacy	−0.03	0.07	−0.51	0.612	−0.16	0.10
Emotional self efficacy	0.08	0.07	1.14	0.255	−0.06	0.21
Persistence	−0.19	0.05	−3.65	<0.001***	−0.28	0.09
Covariance
(Academic self-efficacy, social self-efficacy)	0.47	0.06	8.29	<0.001***	0.36	0.58
(Academic self-efficacy, emotional self-efficacy)	0.50	0.05	9.83	<0.001***	0.40	0.60
(Social self-efficacy, emotional self-efficacy)	0.51	0.05	9.75	<0.001***	0.41	0.61
(Depression, anxiety)	0.80	0.03	24.84	<0.001***	0.73	0.86
(Depression, externalising symptoms)	0.46	0.06	7.28	<0.001***	0.34	0.59
(Anxiety, externalising symptoms)	0.46	0.08	5.59	<0.001***	0.30	0.62

Note. *p < 0.05 **p < 0.01 ***p < 0.001; All measurement model paths were significant at the p < 0.001 level; ¹Robust Standard Errors reported.

Further analysis sought to identify the proportion of the total effect of growth mindset and persistence mediated by self-efficacy on each outcome (depression, anxiety and externalising behaviours) by sex (Table 9).

Table 9. Proportion of indirect effect to total effect mediated by self-efficacy on outcomes by sex.

	Boys			Girls
	Depression	Anxiety	Externalising behaviours	Depression	Anxiety	Externalising behaviours
Growth mindset	1.00	1.00	1.00	0.20***	1.00	1.00***
Persistence	0.14	1.00	0.29***	1.00	1.00	1.00***

Note: *p < 0.05 **p < 0.01 ***p < 0.001.

Results indicated that for boys, growth mindset was fully mediated by self-efficacy for all outcomes. Results comparing the indirect to total effect of persistence on outcomes indicated 14% of the total effect of persistence on depression, 100% of the total effect on anxiety and 29% of the total effect on externalising behaviours was mediated by self-efficacy. For girls, self-efficacy fully mediated the effect of persistence on all outcomes. For growth mindset, self-efficacy mediated 20% of the effect on depression and 100% of the effect on anxiety and externalising behaviours.

Discussion

The primary aim of this study was to explore relationships between motivational mindsets and mental health outcomes and to evaluate the mediating effects of self-efficacy. While previous studies in the United States, Russia, China, and Germany have evaluated the mediating effect of self-efficacy on psychopathology, these studies focus on the mediating effect of self-efficacy resulting from experience of risk factors (e.g. stress and adverse life experiences), finding that higher levels of self-efficacy reduced symptoms of psychopathology (Maciejewski et al., 2000; Schönfeld et al., 2016). In contrast, the focus of the present study was to evaluate the mediating effect of self-efficacy resulting from protective factors, specifically motivational mindsets. Correlation between persistence and growth mindset in this study (r = 0.38) were similar to correlations found between measurement of grit and growth mindset in previous research (r = 0.27), reinforcing the association between these motivational mindsets (Sigmundsson et al., 2020). Evidence evaluating the protective impact of positive characteristics is important because these factors may be more easily modified than risk factors that may be difficult to change (e.g. structural poverty) or impossible to modify if they have already been experienced.

Cultural context, and gender role expectations, have been shown to be key factors to consider when understanding psychological risk and protective factors for anxiety, depression, and externalising symptoms. While evidence suggests the positive effect of self-efficacy on reduced psychopathology in several contexts, additional evidence is needed from diverse contexts to understand how perceived self-efficacy functions as a mechanistic pathway for mental health in different cultures. Researchers posit that differences in perceived self-efficacy may be different in more collectivistic versus individualistic cultures (Bond, 1991). Evidence from China suggest that individuals report lower self-efficacy in comparison to more individualistic cultures (Schönfeld et al., 2016; Schwarzer et al., 1997).

Results from this study indicated differential associations by gender between growth mindset and persistence, mediated by dimensions of self-efficacy on mental health outcomes. In this sample, girls reported lower emotional self-efficacy than boys, and there were no significant differences by sex in academic self-efficacy or social self-efficacy. These finding align with previous research indicating that girls report lower emotional self-efficacy than boys, but have similar academic self-efficacy as boys (Bacchini & Magliulo, 2003; Muris, 2001). In Tanzania, early adolescent girls have reported greater household responsibilities in comparison to boys and less freedom outside the home which may limit coping strategies such as use of social support systems for emotional self-regulation (Cherewick et al., 2021). Gender differences in coping strategies have been explored in other east African cultures that indicate that girls have less freedom to engage with friends and community members outside the home in comparison to boys (Cherewick et al., 2015).

Results indicated that there were no gender differences in internalising or externalising symptoms. While much research has indicated that males typically exhibit more externalising symptoms than girls, gender differences in internalising and externalising symptoms emerge in mid- to late-adolescence (Bongers et al., 2004; Daughters et al., 2009; Hankin et al., 2007; Leadbeater et al., 1999; Rudolph, 2002; Scaramella et al., 1999; Telzer & Fuligni, 2013). In contrast, during early adolescence, differences in psychopathological symptoms by gender have been shown to be insignificant, highlighting early adolescence as a window of opportunity to target modifiable protective factors such as self-efficacy that are precursors to internalising and externalising problems (Rocchino et al., 2017).

In alignment with previous studies, academic self-efficacy was significantly associated with reduced externalising symptoms in both boys and girls. Academic self-efficacy during adolescence has been shown to predict externalising behaviours (Caprara et al., 2008) and is related to risk for hyperactivity, impulsivity, and distractibility (Demaray & Jenkins, 2011; Rocchino et al., 2017; Valle et al., 2006). A meta-analysis of self-efficacy by gender has shown males held slightly more academic self-efficacy than females; however, this difference was largest in late adolescence, which again highlights the importance of targeting self-efficacy as a modifiable protective factor during early adolescence (Huang, 2013). Academic self-efficacy increases educational attainment but has also been shown to lower depression and increase hope for the future (Bandura & Locke, 2003; Roeser et al., 2002). Early adolescent interventions should consider schools as a context to promote academic self-efficacy through increased student engagement, afterschool activities, and career exploration (Alt, 2015; Forrest-Bank & Jenson, 2015; McCoy & Bowen, 2015).

Examination of modification indices indicated that model fit was improved with the addition of a direct effect from growth mindset to depression in girls. In boys, examination of modification indices indicated a direct effect from persistence to externalising behaviours. These direct paths indicate that while self-efficacy mediates a portion of the effect from motivational mindsets on psychological symptoms, these mindsets directly affect depression in girls and externalising behaviours in boys. It is plausible that other psychosocial factors such as self-esteem, autonomy, and social support may also mediate the effect of motivational mindsets on mental health.

Given well-established evidence that girls are especially at risk for internalising disorders (depression and anxiety) than boys and that these differences peak during mid- to late-adolescence (ages 13-18), early adolescence (ages 10-12) is an opportunity to reduce gender disparities in mental health that can persist across the life course (Angold et al., 1998; Beesdo et al., 2009; Beesdo-Baum & Knappe, 2012; Caouette & Guyer, 2014; Salk et al., 2017; Steinberg, 2005). Interventions that aim to increase motivational mindsets should consider integrating components that target academic, social, and emotional self-efficacy. Previous evidence suggest the close association between motivational factors and self-efficacy in predicting academic attainment, internalising and externalising symptoms (Caprara et al., 2008; Huang, 2013). Social emotional learning during adolescence is best supported by opportunities to practice mindsets and skills in different contexts, such as use of social support networks as a coping strategy that could support emotional self-efficacy (Cherewick et al., 2020; Shikai et al., 2007). Interventions targeting self-efficacy should consider implementation methods designed for use in the home, at school, in peer settings and in the community.

Limitations

This study was limited to cross-sectional data in a sample of early adolescents in Dar es Salaam, Tanzania. The results of this study may only generalise to similar populations of early adolescents in low-resource collectivistic contexts. Measures of self-efficacy are self-reported and therefore, require self-appraisal that could bias results if study subject’s current mental health limits capacity for self-appraisal or if subjects are motivated to respond to questions in ways that conform to gender norms and expectations. This study measured persistence; however, future studies should explore other similar constructs such as grit which combines persistence with measures of passion and goal setting. Future studies should examine other potential mediators between growth mindset and persistence on psychopathology such as social capital, social identity and autonomy that have been identified in other studies (Bovier et al., 2004; Häusser et al., 2012; Lai, 2009). Finally, the results presented are correlational and the direction of causality between measured constructs requires longitudinal data. Longitudinal data would also allow for identification of sensitive developmental windows to target modifiable protective factors such as self-efficacy and to explore gender differences in trajectories of mental health.

Conclusion

Findings suggest that self-efficacy is an important mediating factor between motivational mindsets and mental health outcomes for both boys and girls. Consistent with previous research, academic self-efficacy is an important target, particularly for universal prevention programmes seeking to modify protective factors that reduce gender disparities in mental health that amplify during mid- to late-adolescence.

Acknowledgements

The authors would like to thank Health for a Prosperous Nation (HPON) for their work recruiting participants, obtaining consent and assent, and survey administration. Thank you to the Discover National Advisory Board members, Ministry of Education, and Ministry of Health for their support. The authors thank the National Institute of Medical Research in Tanzania (NIMRI) and the University of California Berkeley Committee for Protection of Human Subjects Institutional Review Board for their guidance, approval, and support. Most of all, the authors thank the adolescents and their parents/caregivers for their participation and support in this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are openly available in the CU Anschutz Digital Collections repository at https://digitalcollections.cuanschutz.edu/work/ns/d9f4ed30-c1cb-4495-8508-6c35d65b6a2a.

Additional information

Funding

This work was supported by Bill and Melinda Gates Foundation [grant number OPP1158584].