Exploring voice and agency among adolescents in South Asia: a vignettes-based approach

ABSTRACT The innovative use of vignettes in quantitative surveys offers a novel means of assessing attitudes, capturing individuals’ responses across diverse life domains to measure voice and agency more directly. Utilizing the use of scales and vignette-based inquiries, this study explores voice and agency on making marital decisions among 458 adolescents in India and Nepal. Regression analysis revealed that higher voice and agency scores correlated significantly with reduced odds of endorsing early marriage and increased support for continued education post-marriage. In vignette scenarios, greater voice and agency scores were linked to advocating against early sibling marriage and making independent choices for their own futures. These findings highlight that improved self-perception of agency empowers adolescents to assert independent decisions. Utilizing vignettes proved beneficial in this context as they enabled adolescents to understand community norms through storytelling, offering insights into their likely behaviour in specific situations, rather than drawing on abstract concepts.


Introduction
Adolescents make up 16% of the world population, accounting for 1.3 billion people globally (United Nations Children 's Fund, 2022).Defined by the United Nations as those between the ages of 10 and 19, adolescents experience a significant amount of physical and mental growth, and development during the transition period between childhood and adulthood (United Nations Children 's Fund, 2022).It is also a time when young people begin to understand and form social, cultural, and gender norms (Jaworska & MacQueen, 2015;Sawyer et al., 2018;Vyas et al., 2022).Further, adolescence is an opportunity for growth, exploration, and self-identity (United Nations Children's Fund, n.d..).However, for many adolescent girls, restrictions on freedom of movement, lack of voice and agency, and deeply entrenched norms exacerbate inequities they face, often forcing them into early marriage, early motherhood, increased household work, and gender-based violence (Jewkes & Morrell, 2018;Vyas et al., 2021;Yu et al., 2017).While many programmes are aimed at increasing voice, agency, and gender attitudes, current measures for these outcomes may not always capture the true essence of the constructs being measured.Through the utilization of vignette methodologies, it becomes feasible to quantify the level of agency and vocal expression exhibited by adolescents.This approach facilitates the examination of their responses towards prevailing gender norms characterized by disparities, thereby enabling an appraisal of their capacity to autonomously navigate decision-making processes (Blum et al., 2019).

Voice and agency
Voice and agency are inextricably linked and are key elements of adolescent empowerment, particularly for girls and women.Voice is the capacity of adolescents to speak up and be heard in their communities and share their beliefs around decisions that affect them (Klugman et al., 2014) while agency is the ability of adolescents to make decisions about their life and act on them to achieve their defined outcomes free of any form of violence or fear (Klugman et al., 2014).Certainly, voice is an aspect of agency, since simply articulating needs and interests are not enough to make systematic change.For change to occur, voice must exceed simply speaking up and sharing beliefs.Rather it must lead to agency to act on one's interests (Gammage et al., 2016).By being able to use their voice, adolescents can increase their sense of self-worth and self-efficacy and be able to live their life based on their individual choices (Nagaraj et al., 2022).Further, amplifying the voice and agency of adolescents can lead to practical and long-lasting development for themselves and their communities, explaining why interventions targeting adolescent empowerment utilize these concepts as critical constructs in measuring gender equality (Vyas et al., 2021).

Perceptions of gender norms
Gender norms are culturally prescribed expectations of behaviour and are amongst the strongest factors that influence an individual's attitudes, beliefs, and behaviours (Kågesten et al., 2016;Tolman et al., 2016;Vyas et al., 2021).Like voice and agency, perceptions of gender norms and norms themselves begin to form during adolescence and solidify due to increased social pressures to conform to what is 'appropriate' (Vyas et al., 2021).Further, gender norms are associated with inequitable gender attitudes, which are often perpetuated through substance abuse; violence; and unequal access to resources, power, and education (Kågesten et al., 2016;Vyas et al., 2021).Gender attitudes often refer to an individual's perceptions, beliefs, or support of gender norms (Vyas et al., 2021).Across of every setting, girls and women that are socialized to conform to traditional gender norms often have less voice and agency in making life decisions (Vyas et al., 2021).

Education and early marriage in South Asia
In recent decades, India and Nepal have seen a rapid growth in their adolescent population.India has one of the largest adolescent populations in the world, with every fifth person, or 20% being between 10-19 years of age, accounting for 253 million people (United Nations Children's Fund, n.d..).Similarly, in Nepal, adolescents account for 22% of the population.In both countries, adolescents face vast gender inequities, deeply entrenched norms, lack of safety, and lack of equal access to education and resources, often leading adolescent girls with limited mobility and voice (Vyas et al., 2021).While there has been increased investment in the last decade to improve an adolescent's independent agency, deeply entrenched social norms surrounding marriage and mobility are pervasive.As a result, education is often overlooked for young girls in both countries.A 2015 study conducted in India by Panthhe showed that boys had at least two or more years of schooling compared to their female counterparts, and this gap was wider in rural areas, with boys having three or more years of schooling than girls (Panthhe & Mcchutcheon, 2015).A large part of the educational lag is due to early marriages in both countries.A 2018 census found that 25.3% of girls in India were married by 18, including 5.4% of girls who were married by the age of 15 (World Bank, n.d.).Similarly, in Nepal, the likelihood for early marriage is higher as the 2019 census indicated that 32.8% of Nepalese girls were married by the age of 18, including 7% of whom were married by the age of 15 (World Bank, n.d.).In India, currently the legal age of marriage for women is 18 (India the Special Marriage Act, 1954), while in Nepal it is 20 for both genders (Nepal National Civil (Code) Act, 2017).Gender differences seen in education not only play a role in early marriage, but these differences perpetuate into the labour force and can also exacerbate differences in access and availability of health resources and overall wellbeing of women during their lifespan.

Measuring voice and agency with vignettes
Constructs are typically measured by relying on direct questioning on survey items as has often been seen with respect to voice and agency (Tapal et al., 2017, Vyas et. al;Vyas et al., 2021).However, voice and agency remain difficult constructs to measure, especially with adolescents, and measures have varied widely, which necessitates innovative ways to gauge these constructs in selected communities (Zimmerman et al., 2019).One way to measure voice and agency is to ask adolescents how they would behave and how they would respond in different domains of their life, which would more directly measure the expression of their voice and agency (Klugman et al., 2014) To accomplish this requires the utilization of vignettes.
Vignettes are a research tool to assess attitudes, values, norms, and other sensitive topics in all aspects of public health (Blum et al., 2019).One of the widely used definitions of a vignette is provided by Finch which states vignettes are 'short stories' about hypothetical characters in specified circumstances and the interviewee is incited to respond to that situation (Finch, 1987).A vignette essentially presents a story, or scenario, to the participant and asks how they expect or want the central protagonist of the story to act or respond -they can provide ideas about what the central character should do, should think, or should feel -making them the protagonist proxy (Blum et al., 2019).
Vignettes are useful in research as they help with minimizing socially desirable responses and desensitization of sensitive subject matter and present it to the responder in a simpler more palatable manner (Erfanian et al., 2020).When a vignette is used to elicit a response, the participant is less threatened to speak about their own situation in hypotheticals since they would feel the question itself is non-personal (Erfanian et al., 2020).The non-personal nature of the method is explicated by the fact that it simply produces a snapshot of a situation rather than demand an individual's personalized context to a behaviour.With non-personal information postulated the responder can only look to facts offered in the vignette to make decisions, thus, avoiding nonresponses or responses that may be vague about their own life (Hughes, 1998).In addition to creating a non-personally sensitive mechanism of answering questions, it allows researchers to develop, culturally specific, and tailored questions that most closely resemble their sample (Riley et al., 2021).
The use of vignettes is suggested as an effective tool in obtaining responses for harder to summarize constructs such as perceptions, attitudes and norms especially around more sensitive subjects or subjects that need to be broken down for younger or less literate populations (Erfanian et al., 2020).Vignettes are also a way to identify the validity of scaled constructs.Comparing scaled construct scores of complex items with the responses to vignettes that illustrate cases around those constructs examines if the construct measurements are in the right direction.
Formulating vignettes does not follow a strict linear process.Instead, it is an iterative process that looks at the community context and the research questions and can go through many steps before reaching the ideal vignette for the sample population (Skilling & Stylianides, 2020).Vignettes for this study were created to elicit responses around voice and agency through stories that directly relate to adolescents in the region by employing real-world choices adolescents may have been faced with, or that they might have to make in their near future.

Study purpose
It is evident that increased voice and agency are instrumental in delaying early marriage, increasing access to healthcare, and decreasing violence (Landry et al., 2020;Vyas et al., 2020Vyas et al., , 2021)).This coupled with changing gender attitudes and norms can improve the overall health and wellbeing of women.To accomplish this, constructs such as voice and agency need to be measured but often have not been tested in younger populations or are hard to assess among younger populations with lower literacy abilities.For these reasons, the purpose of this study is to examine the use of vignettes when measuring constructs such as voice and agency.Our aim is to examine associations between vignette responses that amplify positive views on voice and agency with scale-based measures of voice and agency that have been adapted from previously valid and reliable scales utilized in other populations.This analysis will provide a foundation for whether and how vignettes can be used to measure difficult constructs for adolescents in Low and Middle-Income Countries (LMIC)s.

Study instrument
The instrument, including vignettes, was created after a comprehensive formative research phase to identify the dimensions of voice and agency.An exhaustive non-systematic literature review was conducted to gather various definitions for voice, agency, and gender attitudes.After identifying these definitions, in-depth discussions were conducted with partners in India and Nepal that led to selecting and narrowing the dimensions to the value of education, emotional efficacy, freedom of movement, decision-making in the family and community, access to technology, and opportunity for civic engagement and social action.The resulting vignettes include aspects of decision-making, value of education, and emotional efficacy.
After identifying the dimensions, the next step was conducting qualitative research to develop survey items.Four focus groups with 30 girls and 25 boys between the ages of 10 and 16 were conducted in the cities of Delhi and Lucknow in India.Participants were recruited through community partners in India and included adolescents that were enrolled in school and were participants of previous interventions by the community partners.The focus group participants were selected by a convenience sample to ensure diversity in age and gender.
The focus groups were separated by gender and were facilitated in Hindi by a trained member of the research team.
Participants in the focus group were asked a series of open-ended questions to assess their experiences and perceptions about their education and school environment, challenges and relationships with family and peers, future aspirations and challenges, gendered experiences, decision making within schools and families, restrictions and rules within families and schools, and use of mobile technologies.The discussions were recorded, transcribed, and analysed for themes, which found that all the dimensions except those around technology were relevant for the population.
The next stage was the creation of survey scale items, including the vignettes, in English and Hindi and face validity was tested with five girl and five boy adolescents between the ages of 10-14.The final survey included the vignettes around voice and agency assessed in this paper.Further details on creation of the survey and the voice and agency measures are detailed elsewhere (Vyas et al., 2021).

Measures
The voice and agency vignette presented in this paper describes the story of a 15-year-old girl whose parents have arranged her marriage with a 22-year-old man.The vignettes and the associated questions are below-Geeta is 15.Her parents have found her a nice boy, who is 22 and comes from a good family.Geeta's parents have arranged their marriage for next month.Please read each sentence below and tick the box to show how much you agree or disagree with each of them.

Now, imagine if you were Geeta's brother or sister, what would you do? (Only choose ONE answer)
Now, imagine if YOU were Geeta, what would you do? (only choose ONE answer Two questions followed this vignette, the first question consisted of four sub-questions and the second question was multiple-choice.The sub-questions asked participants to respond using a 4-point Likert scale -strongly disagree, disagree, agree, and strongly agree.The second question was a multiple-choice question with four answer choices that asked participants what action they would take if they were the girl's sibling (See Appendix A for vignette, questions, and response options).
The independent variables in this study are the scale measures of voice and agency defined and adapted from the larger study instrument.Thirty three of the survey items specifically asked participants to respond to a series of questions by rating on a 4-point Likert type scale whether they strongly disagree to strongly agree with the survey items (See Appendix B for survey questions around voice and agency).Survey items were newly developed or adapted from publicly available, validated instruments with related constructs around components of voice and agency, as well as gender norms and attributes, such as the Gender Equitable Measurement Scale created for the Gender Equity Movement in Schools programme in Mumbai, India (Achyut et al., 2017), the UNICEF Evaluation of Empowering Young Girls and Women in Maharashtra, India (United Nations Children Fund, 2018), the Gender Equitable Men scale from the Compendium of Gender Scales (Nanda, 2011), and the Girl Rising India Schools Campaign Evaluation (Vyas et al., 2020).

Study sample and data collection
The data for this study were derived from a larger after-school programme and interventions from local community partners.The sample consists of adolescents aged 10-15 years in public schools located in Delhi, India (urban), Uttar Pradesh, India (rural), and Gandaki Province, Nepal (semi-urban).The in-country research staff followed a standardized protocol in administering the survey to the adolescent participants, which was completed via paper and pencil in an afterschool setting with no teachers or other authority figures involved.The survey was developed in English then translated to Hindi/Nepali and back translated to English by two native Hindi and Nepalese speakers to address any language nuances.Prior to the start of data collection, the community partners reached out to the parents of the participants to provide an overview of the project and data collection instruments and provide a copy of the consent.Verbal consent was obtained from the parents for their child to participate by the in-country community partners.Upon survey completion, the incountry research team entered the completed surveys into a pre-populated survey database in Microsoft Excel.The final datasets were securely transferred to the US research team to download, clean, and analyse.The total sample size was 458 adolescents.This study was reviewed and approved by The George Washington University Internal Review Board (IRB #191568).
Table 1 presents the study sample's characteristics.There were more girls than boys (56.11 vs. 43.90)with an average age of 12.41.Much of the sample was in 6 th grade (30.79%), followed closely by 7 th grade (24.24%) and 8 th grade (17.69%).The sample was divided almost evenly between India and Nepal with 30.35% from rural India, 25.76% from urban India, and 43.89% from Nepal.Most respondents had brothers (71.62%) and/or sisters (65.28%).

Statistical analysis
The 4-point Likert-type responses for each of the items related to decisions around early marriage were collapsed into a dichotomous variable (agree = 1 vs. disagree = 0) (See Appendix C for the response distribution for all options before collapsing into dichotomous form).Frequencies were calculated to identify the level of agreement with each of the items related to marriage.The top four identified options for each vignette question and sub-question were further analysed by running logistic regressions with voice and agency scales as the independent variables, controlling for location, and gender.The regression results are reported in the form of adjusted odds ratios All analyses were conducted in Stata SE 15.1, StataCorp LLC, TX.

Results
Most respondents (78.60%), disagree that Geeta's parents made the right decision in arranging the marriage for 15 year old, Geeta.The respondents, overall, agree with behaviours that involve Geeta using her voice and agency to intervene in the parent's decision with 68.34% agreeing that Geeta should voice her concerns to her parents, 70.02% agreeing that Geeta should propose to stay in school after marriage, and only 38.42% agreeing that Geeta should respect her parent's decision (Table 2).
After controlling for geography and gender, respondents with higher scores of voice (aOR = 0.64,95%CI 0.42-0.95) and agency (aOR = 0.63; 95% CI 0.45-0.88)have significantly lower odds of agreeing that Geeta's parents made the right decision.Respondents with higher scores on the construct measured scales for voice (aOR = 1.89; 95% CI 1.30-2.75)and agency (aOR = 1.87; 95% CI 1.36-2.48)had significantly higher odds of agreeing (vs.disagreeing) that Geeta should tell her parents she does not want to get married (Table 3).Similarly, those with higher reported voice (aOR = 2.60; 95% CI 1.74-3.88)and agency (aOR = 1.84; 95% CI 1.33-2.56)had significantly higher odds of agreeing that Geeta should tell her parents that she will get married as long as she stays in school.There were no significant relationships between voice or agency for the other outcome variables, Geeta's parents made the right decision or Geeta should respect the decision her parents made.
Continuing the conversation on voice and agency around marriage decisions, respondents were asked about hypothetical actions they would take if they themselves were Geeta, the character of the vignette.The most popular choice among them was to tell the parent they do not want to get married (31.00%), followed by 24.67% electing to report the situation to the police and 17.03% electing to ask an adult for help (Table 4).19.21% of the sample indicated that they would get married as that is what the parents decided for them, making not acting the 4 th most selected option for the vignette.As seen in Table 5, Respondents with those with higher agency scores showed lower odds (aOR = 0.63; CI 0.44-0.90) of getting married according to their parents' wishes.There was no significant relationship between higher scores of voice and agency and telling parents that they do not want to get married, going to the police or asking an adult for help.
In the next portion of the vignette, respondents were asked about hypothetical actions they would take if they were Geeta's sibling.The most popular selected option was to tell the parent to not get Geeta married, even though it would upset the parent (32.53%).Like the previous section on if they were Geeta, the second most selected option was going to the police (24.67%), followed by to let Geeta get married since that is what the parents want (14.41%), and the 4 th most selected option was to ask an adult for help (12.66%)(Table 6).
As seen in Table 7, Respondents with higher voice (aOR = 1.56, 95% CI 1.05-2.33)and higher agency (aOR = 1.61, 95% CI 1.18,2.16)scores had significantly higher odds of selecting the option of telling their parents not to get Geeta married even if that would upset them.Similarly, those with higher agency scores had lower odds (aOR = 0.52, p 0.35,0.76) of telling Geeta to get married since that is what their parents want.There was no significant relationship between voice and agency and asking an adult for help or going to the police.

Discussion
Traditional measures and scales for voice and agency are often limited for adolescents as they may not be developmentally appropriate; they can be tedious for data collection; and they often lend themselves to social desirability bias (Fisher & Katz, 2000;Theriault & Holmberg, 1998).This paper examined how a vignette focused on early marriage with constructs of voice and agency embedded in the response options can be applied as an effective measurement tool with adolescents in South Asia.The vignette was created by adolescents themselves when they shared their own lived experiences during the qualitative phase of the study (Nagaraj et al., 2019), and therefore the vignette was contextually relevant and reflects the realities of what many adolescents in South Asia face.Studies that have examined the use of vignettes with adolescents continue to be sparse in the literature (Blum et al., 2019), and this study is one of the first to not only present how adolescents responded to a vignette, but also how the various response options were correlated with traditional scales that measure voice and agency.
The vignette examined in this study had multiple response options with varying levels of voice and agency reflected in each.Our analysis found that vignette response options that were centred around speaking up and making decisions to not engage in early marriage were highly correlated with traditional measures of voice and agency scales, lending credibility and validity to the vignette as a method of measuring voice and agency as it related specifically to early marriage.For example, response options that centred around speaking up to parents, going to the police, or asking other adults for help were all correlated with higher levels of voice and agency as measured with traditional scales among the same group of adolescents.And response options that centred around obeying parents or staying silent were correlated with lower levels of voice and agency measured with traditional scales.This study's findings demonstrate that adolescents can successfully and honestly respond to vignettes, and the insights gained from this analysis support the feasibility of vignettes as a measurement tool.Vignettes are interesting, interactive, and engage adolescents in what is often an onerous and daunting task with respect to survey data collection.Vignettes also minimize response bias as they engage young people in the stories and ask them to respond to questions in a way that transports themselves into the characters in the vignette.Further, the vignette allowed us to examine how adolescents would respond to the story from the perspective of different 'characters.' Few studies have been conducted on the use of vignettes in low and middle-income countries, and therefore this research and instrument fills an important gap.It will be important for future studies to explore the use of vignettes as a measurement tool with adolescents, and to examine whether responses correlate with traditional measures.
There are limitations to this study.First, the formative phase did not include a rigorous systematic review of the literature, nor did it include the development of a conceptual/theoretical framework for voice and agency.Instead, the initial phase focused on qualitative research and thematic analysis to build upon previous studies of voice and agency (Nagaraj et al., 2019).Second, the study only included adolescents enrolled in school and therefore these findings cannot be generalized to adolescents who are out of school.Given the unique needs and context for adolescents, particularly girls, and the gender discrimination they face, it is plausible that out-of-school adolescents' voice and agency significantly differs from young people in school.Furthermore, conceptions pertaining to voice and agency manifest notable variations contingent upon the socio-economic strata and cultural beliefs of both parental figures and adolescents.Consequently, future investigations could leverage the vignette methodology to undertake a more nuanced and detailed examination of these dynamics.
An additional potential limitation might stem from the limited sample size; nevertheless, it is worth noting that the utilization of vignette techniques in research remains comparatively new, resulting in a lack of analogous studies characterized by larger cohorts.Subsequent research could thus include more expansive and heterogeneous samples to empirically test the efficacy of this methodological approach.Despite these limitations, the strengths of the study discussed above cannot be overstated as the findings fill a critical gap in the measurement of voice and agency among adolescents.Further, the results provide new insights to guide future research and measurement development.Most importantly, the surge of new programmes focused on gender sensitization and empowerment can operationalize the vignette presented in this manuscript to appropriately test outcomes and impact.

Table 2 .
Agreement (%) around decisions presented around marriage norms in the vignette.
Do you agree or disagree that Geeta should respect the decision her parents made?176 38.42 282 61.57

Table 3 .
Logistic regression models for vignette response options.

Table 5 .
Logistic regression of ranked options, and voice and agency.

Table 4 .
Ranked (%) responses to imagine YOU were Geeta.What would you do?

Table 6 .
Ranked (%) responses to imagine you were Geeta's SIBLING.What would you do?

Table 7 .
Logistic regression of ranked options, and voice and agency.