Introduction

Within psychology, resilience has been defined and conceptualized in a variety of ways, such as a trait (e.g., Connor & Davidson, 2003), a process (e.g., Luthar et al., 2000) and an outcome (e.g., Smith et al., 2008). While some scholars have described it as a fixed individual trait (e.g., Wagnild, 2009), others have conceptualized it as a dynamic process (Masten, 2001), that is characterized by continuous growth and adaptation (Gartland, 2009). Other definitions, while emphasizing its changing nature, also draw attention to the complex interplay of personal and contextual factors through which an individual draws resources from their environment and its nested levels including family, friends, school and community for navigating through life’s adversities (Olsson et al., 2003). This understanding of resilience is based on the ecological–transactional model of Cicchetti and Lynch (1993).

Of particular importance is the need for the systematic examination and understanding of such resilience factors among adolescents, as half of the mental health issues have an early onset of about 14 years (WHO, 2019). The protective and promotive role of resilience for the adolescent population is evident from studies, showing that adolescents with higher levels of resilience are better able to cope with adversities in adulthood (Werner, 1995), and are also better able to avoid risk-taking behaviors including substance abuse (Resnick, 2000). Studies have also highlighted the link between resilience and positive youth development (Lee et al., 2013) as well as shown resilience to be a positive predictor of academic performance (Kwek et al., 2013). Moreover, resilience has been found to have a strong and consistent relation with well-being in the adolescent population (Faisal & Mathai, 2017).

It is considered to be a “key individual characteristic in the well-being of individuals” and has a positive association not only with hedonic but also eudaimonic well-being (Di Fabio & Palazzeschi, 2015, p. 2). Within the well-being literature, hedonia and eudaimonia comprises a dominant framework for conceptualizing and understanding well-being. The hedonic tradition focuses on attaining pleasure and avoiding pain, whereas the eudaimonic tradition focuses on meaning and personal fulfillment (Huta & Waterman, 2014). Ryff’s model of psychological well-being (Ryff, 2013), for example, is based on a eudaimonic perspective of well-being, while Diener’s conceptualization of subjective well-being is based on a hedonic framework (Diener, 1984). Some other scholars such as Tennant et al. (2007)  incorporated both these approaches in their measure of well-being named the Warwick–Edinburgh Mental Well-being Scale. Additionally, there are also studies that have used separate measures for assessing hedonic and eudaimonic well-being (Jongbloed, 2018). Interestingly, the literature on both well-being and resilience are characterized by a lack of consensus regarding their measurement and definitions (Schultze-Lutter et al., 2016).

As mentioned previously, resilience has been defined and operationalized variously—as a trait, a process and an outcome. Although majority of the early measures (e.g., Resilience Skills and Ability Scale; Jew et al., 1999; Adolescents Resilience Scale; Oshio et al., 2003) treated resilience as an individual trait or included only specific environmental aspects (Friborg et al., 2003), the more recent ones assess and highlight its multidimensional nature (Ungar, 2014) and operationalize resilience as a dynamic process. The Adolescent Resilience Questionnaire (ARQ; Gartland et al., 2011) is one such measure that was developed specifically for the 11- to 19-year age group. Its strength lies in having a strong theoretical grounding as it is based on the ecological–transactional model (Cicchetti & Lynch, 1993). It encompasses diverse individual (e.g., confidence, negative emotions and others) and environmental factors (e.g., neighborhood, school setting, family and community) that mutually interact during adolescence, and shape resilience outcomes (Gartland et al., 2011), thus providing a multidimensional and comprehensive understanding of resilience.

The ARQ comprises 88 items and measures resilience across five domains, including individual, peer, community, school and family, and has 12 scales. Focus group discussions and an in-depth examination of the resilience literature formed the basis of the development of the scale items. While it was developed and validated in the Australian setting, its psychometric properties and factor structure have been investigated across different cultural contexts, including Nepal (Singh et al., 2019), Iran (Cheraghi et al., 2017), and Spain (Guilera et al., 2015). Although these studies confirmed the factorial validity of the ARQ in their particular cultural contexts, certain culture-specific modifications and recommendations were made in some of them (e.g., Cheraghi et al., 2017). The American validation (Anderson et al., 2020), on the other hand, could not confirm its original factor structure, thus highlighting the need for examining its psychometric properties in varied cultural contexts.

Despite being an invaluable tool for researchers, practitioners and other stakeholders of adolescents’ healthy development, the ARQ has not been validated in India. Currently, India is home to the largest proportion of the adolescent population in the world (Sivagurunathan et al., 2015), and the empirical examination and understanding of their resilience will have significant implications. It will enable researchers to better understand its risk and promotive factors among Indian adolescents and how to promote or address them. This necessitates the examination of the psychometric properties and validation of the English as well as Hindi versions (43.63% of the Indian population understands and speaks in Hindi; Census, 2011) of the ARQ in the Indian context.

Existing studies in other cultures have also examined the link between resilience and a diverse range of socio-demographic variables including age and gender (Nourian et al., 2016), and type of schooling and area of residence (Singh et al., 2019). However, their findings are inconsistent. For instance, some studies found females to be more resilient than males (Nourian et al., 2016), while others found the opposite (Moksnes & Eliertsen, 2015). A few other studies did not find any significant gender difference (Lundman et al., 2007). Similarly, while some studies (e.g., R. Singh et al., 2019) found private school-going adolescents to score lower on resilience as compared to public school-going adolescents, others did not find any significant relationship between the type of school and adolescents’ resilience levels (Zuill, 2016). In the case of age, some studies report primary-grade adolescents as being more resilient than middle and higher-school adolescents (Banerjee et al., 2018), whereas others report high-school adolescents as being more resilient than their younger counterparts (Nourian et al., 2016). While these inconsistent findings warrant further investigation, examining the demographic correlates of resilience in particular cultural contexts can also help scholars and practitioners understand the needs of different groups of adolescents. Thus, enabling them to make timely and pertinent policy recommendations. Moreover, using tools such as the ARQ will help researchers to examine adolescents’ resilience across varied ecological domains, thus providing a more fine-grained understanding. However, there is a dearth of such studies in India.

Accordingly, the present study had a twofold objective:

  1. (a)

    To assess the psychometric properties and factor structure of the English and Hindi translated version of the ARQ in India, and

  2. (b)

    To examine the role of select socio-demographic variables (age, gender, place of residence and type of schooling) on the resilience levels of Indian adolescents.

Methods

Participants

The study sample consisted of 1290 adolescents aged between 12 and 18 years (M = 15.27, SD = 1.08) studying in grades five to twelve. Among them, 52.09% attended government schools and 47.90% attended private schools (0.15% did not report their school type), 50.38% were males, and 49.6% were females (0.15% did not report their gender), 55.1% resided in urban settings, and 44.89% resided in rural settings (5.89% did not report their area of residence), and 52% of the participants responded to the data booklet in English, while 47.9% responded in Hindi.

Measures

Adolescent Resilience Questionnaire (ARQ; Gartland et al., 2011)

This 88-item measure comprises 61 positively and 27 negatively worded items. It assesses the resilience of 11- to 19-year-olds across five domains, viz. individual, family, peers, school and community. The individual domain has five factors, while the family, peer and school domains have two factors each, and the community domain has one factor. Thus, there are a total of 12 sub-factors. The scale items are rated on a five-point Likert scale (1 = Never and 5 = All the time) and a high score indicates a higher level of resilience and vice-versa. The Cronbach’s alpha reliability of the 12 scales lies within the range of 0.70 to 0.90 (Gartland et al., 2011).

Warwick–Edinburgh Mental Well-being Scale (WEMWBS; Tennant et al., 2007)

The WEMWBS is a 14-item measure of positive mental health that evaluates both hedonic and eudaimonic well-being of the respondent over the past two weeks. It comprises positively worded items only and is rated on a five-point Likert scale (1 = none of the time to 5 = All of the time). While the internal consistency reliability of the original scale was found to be 0.89, its Indian validation yielded a Cronbach’s alpha value of 0.84 (Singh & Raina, 2020).

Procedure

Various government and private schools situated in different parts of Delhi NCR, Haryana, Rajasthan and Uttar Pradesh were approached for data collection. Upon explaining the purpose of the study and obtaining the consent of the school authorities, an introductory session was organized. This session was attended by those students of grades five to twelve who were willing to participate, and their class teachers. The teachers also facilitated the classroom data collection process. Firstly, the interested students were explained the aim of the study and assured of the confidentiality and anonymity of their responses, as well as the voluntary nature of their research participation. It was also emphasized that they could withdraw their participation at any stage. Detailed verbal instructions were given to the participants for responding to the study measures, the consent form and the demographic information sheet (age, sex, grade, area of residence, school type). Further, they were given the option of responding to the booklet either in Hindi or in English, depending upon their comfort and preferred language. The measures had been translated to Hindi for this purpose, following the standard forward and backward translation procedure (Beaton et al., 2000).

Data Analysis

As the first step for data analysis, missing value analysis was done using SPSS version 20. The missing data were reported to be below 5% and were imputed using series mean imputation. Thereafter, descriptive statistics, item analysis, exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) were conducted to assess the psychometric properties of the ARQ in the Indian context. For examining the role of socio-demographic factors on resilience, multivariate analyses were conducted and effect sizes (partial eta squared) were estimated following the guidelines of Cohen (1988) where 0.01, 0.06 and 0.14 represent small, medium and large effects, respectively. SPSS was used for descriptive and multivariate analyses, item analysis and EFA, and LISREL version 8.8 was used for CFA.

Results

In line with the primary objective of this study, we firstly set out to confirm the factor structure of the ARQ in English and in Hindi. However, as the CFA models did not converge, we decided to conduct EFA on the data of the English respondents (n = 671) after doing an item analysis. Upon exploring and finalizing the factor solution, item analysis was done again to check the properties of the finalized items. The next step involved validating the factor structure of the finalized scale on the data of the Hindi respondents (n = 619).

Psychometric Properties of the Brief ARQ

Item Analysis and Exploratory Factor Analysis (EFA)

Item analysis of the English data set showed that the mean values for the items ranged from 1.83 to 4.29 (SD = 0.83 to 0.85). The skewness and kurtosis (skewness = − 1.10 to 0.83; kurtosis = − 1.10 to 1.19) were within the acceptable range of below 2.0 for skewness and below 7.0 for kurtosis (Curran et al., 1996), thus demonstrating the normal distribution of scores. However, based on low corrected item-total correlation (values below 0.20; Field, 2005), 48 items were deleted. Thus, a total of 40 items were retained after item analysis.

As the Bartlett’s test of sphericity was significant (p < 0.001), and the Kaiser–Meyer–Olkin (KMO) values (0.77 to 0.89) were above the cut-off point of 0.60 (Kaiser, 1974), we proceeded with EFA using maximum likelihood extraction. After exploring several factor-solutions, we finalized a one-factor structure for each of the five domains of the ARQ (i.e., individual, family, peer, school and community). The resulting scale had 35 items with factor loadings ranging between 0.40 and 0.77 (Hinkin, 1995, 1998) (see Table 1).

Table 1 Final items of the Brief 35-item ARQ
Full size table

The mean value of these 35 items ranged from 3.25 to 4.23 (SD = 0.86 to 1.14), skewness from − 1.10 to − 0.25, kurtosis from − 0.71 to 1.19 and corrected item-total correlation values from 0.3 to 0.54. Thus, indicating acceptable properties based on the norms recommended for item analysis in the extant literature (Curran et al., 1996; Field, 2005).

In sum, the resulting measure had five factors (or “domains” as Gartland et al., 2011; p.4 called it), each having a single-factor structure, and a total of 35 items.

Confirmatory Factor Analysis (CFA)

The next step was to confirm the new factor structure of the ARQ on the data of the Hindi respondents (n = 619). The resulting CFA showed that the model fit the data to an acceptable level for the Hindi version. The goodness-of-fit indices, their norms and the obtained values are mentioned in Table 2.

Table 2 The results of the confirmatory factor analysis of the Brief 35-item ARQ
Full size table

Reliability, Validity, Floor and Ceiling Effect

The Cronbach’s alpha of the Brief ARQ was found to be 0.88 and the reliability values of its sub-scales ranged between 0.66 and 0.82. Both the English and Hindi versions of the Brief ARQ had a Cronbach’s alpha value of 0.89, thus demonstrating its adequate reliability. Further, the significant positive correlations of all the five domains of the Brief ARQ with the WEMWBS, which is a measure of well-being, established its convergent validity (see Table 3 for details).

Table 3 Descriptive statistics and convergent validity
Full size table

The final scores on the Brief ARQ were also assessed to see if there were any problems related to floor and ceiling effect. Based on the recommendations in the existing literature (Terwee et al., 2007; Windle et al., 2011), it was decided that if more than 15% of the participants obtained the highest or the lowest possible scores that would be considered as a problematic floor and ceiling effect. However, no floor or ceiling effects were observed as none of the respondents obtained the lowest (35) or the highest possible score (175).

Resilience and Demographic Variables

The relationship between demographic variables, and the Brief ARQ and its domains were examined through multivariate analysis. This assessment was done on the data of 1205 respondents, as 85 of them did not report various demographic information.

As may be observed from Table 4, a significant main effect was observed for all the demographic variables, namely age (early vs. late adolescence;12–15 years; mean = 131.74 and 16–18 years; mean = 135.69), sex (male; mean = 131.88 vs. female; mean = 134.69), type of school (government; mean 135.23 = vs. private; mean = 131.18) and area of residence (rural; mean = 134.44 vs urban; mean = 131.84), as well as for the interaction of age and area of residence, and school type and area of residence, and the three-way interaction between age, school type and area of residence, and sex, school type and area of residence. However, all their effect sizes were small.

Table 4 Main effect of the demographic variables on resilience
Full size table

With regard to the different domains of resilience (see Table 5), the multivariate results revealed that adolescents aged 16–18 years obtained higher scores (p < 0.05) on three resilience domains, namely individual (M = 24.27), family (M = 40.45), peer (M = 27) and school (M = 27.20) domains vis-a-vis their younger counterparts (individual M = 22.93; family M = 39.46; peer M = 25.77; school M = 26.67). The effect size of the differences was small.

Table 5 Effect of demographic variables on the five resilience domains
Full size table

In terms of gender, females obtained higher scores (p < 0.05) than males on the family (females; M = 40.49 vs males; M = 39.21), peer (females; M = 26.80 vs. males; M = 25.69) and school (females; M = 27.24 vs males; M = 26.51) resilience domains. The effect size of the differences was small for family and peer domains, and not substantial for the school domain.

With regard to type of school, adolescents studying in government schools scored higher (p < 0.05) than adolescents studying in private schools in individual (govt school; M = 24.22 vs private school; M = 22.63), peer (govt school; M = 26.93 vs private school; M = 25.51), school (govt school; M = 27.28 vs. private school; M = 26.45) and community (govt school; M = 17.13 vs private school; M = 16.56) domains. The effect size of the differences was small.

Finally, while the rural adolescents had higher mean scores (p < 0.05) on peer (rural; M = 26.74 vs urban; M = 25.64), school (rural; M = 27.13 vs urban; M = 26.56) and community domains (rural; M = 17.71 vs urban; M = 15.79), the urban adolescents scored higher in the family domain (urban; M = 40.57 vs rural; M = 39.27). The effect size of the difference was small for family, peer and community domains, and not substantial for school resilience domain.

Discussion

To the best of our knowledge, this was the first study examining the psychometric properties and factorial validity of the ARQ in the Indian socio-cultural milieu. The original ARQ developed in the Australian context has 88 items, five domains or factors (viz. individual, family, peer, school and community) and 12 sub-factors. The Indian validation found a reduced 35-item version with five domains having a single-factor structure each, as being the most adequate fit in the Indian context. Similar to our findings, Anderson et al. (2020) failed to confirm the original factor structure of the ARQ in a sample of 3222 adolescents and found a reduced 49-item version of the ARQ to have a better fit and more acceptable psychometric properties in the American context.

Although some other validation studies of the ARQ in other cultural contexts (Cheraghi et al., 2017; Guilera et al., 2015; Singh et al., 2019) confirmed its original factor structure, certain culture-specific modifications were either incorporated or recommended by their authors. For instance, the Persian validation resulted in the exclusion of one item from the community domain. Further, their obtained model was a fair fit only, and the authors attributed this to the poor performance of the community domain, where all the items had low factor loadings (0.12 to 0.24) with the exception of one item (0.60). Hence, they recommended a significant revision of the community scale in the Iranian context (Cheraghi et al., 2017). Similarly, the authors of the Spanish validation (Guilera et al., 2015) recommended a revision of the empathy/tolerance scale of the individual domain on account of its low reliability and low corrected item-total correlation. The Nepalese validation study, on the other hand, reported a lower alpha reliability value of the confidence scale (individual domain) as compared to its reliability in the original study. Based on this, the authors recommended rewording the items of the confidence scale in the Nepalese version (Singh et al., 2019).

In the present study, it was observed that the Brief 35-item ARQ had good reliability (α = 0.88) for both English (α = 0.89) and Hindi versions (α = 0.89), and did not demonstrate any floor or ceiling effects. Moreover, consistent with existing studies (e.g., Liebenberg & Moore, 2018; Patil & Adsul, 2017; Smith & Hollinger-Smith, 2015; Singh et al., 2018), all the five domains of resilience were found to have a significant positive association with well-being, thereby establishing the convergent validity of this resilience measure.

Moving on, the results of the multivariate analysis showed a significant main effect for all the socio-demographic variables, namely age, sex, area of residence and type of schooling. Existing studies on the adolescent population also report a significant main effect for gender (Liebenberg et al., 2012), and a significant association of resilience with their age (Nourian et al., 2016), area of residence and type of schooling (R. Singh et al., 2019).

The present results further showed that older adolescents, those aged between 16 and 18 years obtained higher scores than their younger counterparts aged between 12 and 15 years in various resilience domains, namely individual, family, peer and school. Some other studies have also found similar results (Nourian et al., 2016; Singh et al., 2020). However, in a few other studies younger adolescents obtained higher resilience scores than older adolescents (Banerjee et al., 2018), thus pointing to an inconsistency in this finding in the existing literature.

The present results also show that government school-going adolescents obtained higher scores than private school-going adolescents in different resilience domains (individual, school and community resilience domains). In India, private schools are regarded as providing better amenities and facilities than government schools (Gouda et al., 2013), and private school-going adolescents have reported better levels of psychological well-being, physical health, quality of life and even environmental conditions (e.g., Singh & Junnarkar, 2014). Interestingly, resilience levels seem to show the opposite trend not only in the present but also in previous studies (Choudhury & Sharma, 2019). This warrants further investigation.

Our findings also indicated that females obtained higher scores vis-à-vis their male counterparts in various resilience domains and this is consistent with the existing evidence in the adolescent resilience literature (Nourian et al., 2016). Finally, rural adolescents of the present study obtained higher resilience scores than urban adolescents in the domains of community, peer and school. This seems to be in line with existing findings which shed light on the socially inclusive, stable and supportive environment of the rural settings (Crowell et al., 1986), along with the higher levels of well-being and sense of belongingness of rural adolescents vis-à-vis their urban counterparts (Freeman et al., 2001, in Singh & Raina, 2020). In contrast to these ecological domains, urban adolescents obtained higher scores than rural adolescents in the family domain. Indeed, existing studies show that urban adolescents perceive their family environment to be better than rural adolescents. They reported greater independence, greater freedom of thought and expression, and perceive their family to be more caring and accepting (Balda et al., 2019).

Notably, although we found a significant main effect for all the socio-demographic factors, namely age, sex, type of school and area of residence, their effect sizes were small. Likewise, a few existing studies (e.g., Anderson et al., 2020; Liebenberg et al., 2012) have also noted small differences in resilience between adolescent males and females. Yet others have reported a significant main effect for both gender and age (Sun & Stewart, 2007), without documenting their effect sizes. Although the p value helps in understanding whether an effect exists, it cannot reveal the size of the effect, and scholars are increasingly pointing to the value of examining and reporting effect size (Sullivan & Feinn, 2012). The results of the present multivariate analysis are a case in point, and our obtained p values should be considered in light of their effect sizes.

Besides, our study findings should also be considered in light of its limitations, namely that the present data were collected only from selected regions of northern India, and from English and Hindi-speaking school-going adolescents. Future research in India may not only benefit from including participants from diverse backgrounds and regions of the country but also from examining the role of other socio-demographic factors including family education, family income, family type and size, on the resilience levels of adolescents.

By providing a brief, validated and culturally appropriate version of both the original English, and Hindi translated versions of the ARQ in the Indian context, the present research has widened the scope for future resilience research on the adolescent population in India. This is particularly important considering that Hindi is one of the most widely spoken and understood languages in our country (Census, 2011). Moreover, given its strong theoretical grounding and multidimensional approach to resilience, the Brief 35-item ARQ is an invaluable tool for Indian researchers and other stakeholders of adolescents’ well-being to assess their resilience levels across varied individual and ecological domains.