(2015) Initiative pour les adolescentes en Haïti (AGI) : Rapport de projet

April 2015 Document of the World Bank Haiti Adolescent Girl Initiative (AGI): Project Report Poverty Global Practice Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized 1 Contents ACKNOWLEDGEMENT .............................................................................................................................................. 3 ABBREVIATIONS AND ACRONYMS ...................................................................................................................... 4 SUMMARY ......................................................................................................................................................................... 5 I. Introduction ................................................................................................................................................................ 8 II. Framework of Analysis: Expanded Assets Approach to Supporting Employability ...................................... 9 III. The Haiti AGI project ........................................................................................................................................ 13 IV. Who are the AGI trainees? ................................................................................................................................ 22 V. Impacts on overall agency and empowerment .................................................................................................... 24 VI. Employment-related impacts in the short term and expected medium term ............................................ 30 VII. Considerations for future programs ................................................................................................................. 33 Integrated design of the project ............................................................................................................ 33 Complementing technical training with socio-emotional and soft-skills development can be a cost- effective program component to enhance results .................................................................................. 34 Aspirations and post-training trajectories .............................................................................................. 36 Job search experience ........................................................................................................................... 39 Cost-effectiveness ................................................................................................................................ 41 Conclusion ......................................................................................................................................................................... 44 References .......................................................................................................................................................................... 47 Annex X.1: Summary of main components of the Haiti AGI .................................................................................. 51 Annex X.2: Haiti AGI timeline ....................................................................................................................................... 53 Annex X.3: Main labor characteristics of Haitian youth (17-21) – ECVMAS ........................................................ 56 Annex X.4 Reframing Employability for Poverty and Shared Prosperity ............................................................... 57 Annex X.5: Full list of SMS job search motivation messages ................................................................................... 61 Annex X.6: Full list of SMS job search survey questions ........................................................................................... 67 Annex X.7. Press release of the Graduation of the First Cohort of the Haiti AGI (June 2013) ......................... 68 Annex X.8: Mid-line Qualitative Evaluation (May 2013) ......................................................................................... 70 Questionnaire for the participants/ beneficiaries of the first cohort.......................................................... 73 Questionnaire for the project coordinator at community NGOs and training centers ............................... 75 Questionnaire for the parents of Haiti AGI participants ........................................................................... 77 Questionnaire for mentors ....................................................................................................................... 79 Annex X. 9: End-line Qualitative Evaluation (Dec. 2014 –January 2015) .............................................................. 81 Annex X.10: Identified research gaps based on the Haiti AGI experience ............................................................. 84 2 Annex X.11: Impact evaluation methodology and results ......................................................................................... 86 Methodology ............................................................................................................................................ 86 Research design .................................................................................................................................... 86 Sample and attrition ............................................................................................................................. 87 Identification ........................................................................................................................................ 88 Baseline characteristics and balance between treatment and control groups at baseline ......................... 89 Results of Balance and Attrition ............................................................................................................... 91 Characteristics and balance of the population with and without attrition .................................................. 92 Baseline characteristics and balance between treatment and control groups .............................................. 99 Impact Results ....................................................................................................................................... 108 List of Figures and Tables Figure 1. Integrating the asset and the capability approach: The endowment, use and return of assets .......... 10 Figure 2. Breaking down agency and socio-emotional capital ........................................................................ 11 Figure 3: The Haiti AGI integrated project design ........................................................................................ 13 Figure 4. Impact of the AGI project on autonomy in decision making ....................................................... 25 Figure 5. Impact of the AGI project on decisions related to personal mobility ............................................. 25 Figure 6. Impact of the AGI project on the support network of the beneficiaries ......................................... 27 Figure 7. Impact of the AGI project on attitudes towards violence with partner ...................................... 28 Figure 8. Impact of the AGI project on agency related outcomes .............................................................. 29 Figure 9: Word mapping of the results of the First Qualitative Evaluation (May 2013) ................................. 30 Figure 10. Impact of the AGI project on labor outcomes ........................................................................... 30 Figure 11. Impact of AGI on expectations about jobs, income and education .............................................. 32 Figure 12 : Aspirations are shaped by both internal and external factors. ...................................................... 36 Figure 13: Typology of AGI trainees' educational and professional aspirations and their trajectories ............ 38 Figure 14: Haiti AGI cost shares per project component ............................................................................ 42 Figure 15: Shares of training and stipend costs per beneficiary (USD 1,444)............................................. 42 Box 1: Using SMS-based encouragement to foster agency in labor market transition ........................ 40 Table 1: Training cost comparison of AGI pilots ......................................................................................... 41 Table 2: Examples of quotes collected as part of the focus groups ........................................................... 71 3 ACKNOWLEDGEMENT This report summarizes the results of the implementation and evaluation of the Haiti Adolescent Girl Initiative project (P123483). The project and evaluations were led by Aude-Sophie Rodella and Facundo Cuevas (GP- Poverty). The report was written by Aude-Sophie Rodella, Facundo Cuevas and Bernardo Atuesta with the support of Carolina Ferrer-Rincón and Thiago Scot. On the Bank side, the Haiti AGI team included: Aude-Sophie Rodella (TTL), Facundo Cuevas (co-TTL), Sandra Jean-Gilles (Field Coordinator), Maria-Beatriz Orlando (former TTL- until 2012), Bernardo Atuesta (IE and M&E), Carolina Ferrer-Rincón (Operational support), Rebecca Schutte (Operational support and qualitative evaluation), Sarah Haddock (Operational support) and Atwood Raphael (M&E). Over the course of the project, administrative support was provided by: Karina Brito, Sandra Milord, Ane Perez-Orsi, Tatianna Guerrante, Amparo Lezama-Manta, Karem Edwards. The project was implemented under the supervision of Louise Cord (Practice Manager, LAC Poverty GP), Mary Barton-Docks (Special Envoy for Haiti), Alexandre Abrantes (Former Special Envoy to Haiti) and Raju Singh (Program Leader, Haiti). The team benefited from valuable advice and support throughout the design, implementation and evaluation of the project notably from Michelle Keane, Maria-Beatriz Orlando, Sarah Nedolast, Luc Razafimandimby, Elizabeth Ruppert, Pierella Paci, Sylvie Debomy and Jessica Terry. 4 ABBREVIATIONS AND ACRONYMS AGI: Adolescent Girl Initiative CPFPC : Centre Polyvalent de Formation Professionnelle de Carrefour - Multi-purpose technical training center of Carrefour ECD: Early Childhood Development ECVMAS: Enquête sur les Conditions de Vie des Ménages après le Séisme EOI: Expression of Interest IE: Impact Evaluation IGA: Income-Generating Activity INFP : Institut National pour la Formation Professionnelle et Technique - National Institute for Technical Training GAP: Gender Action Plan MEFP: Ministère de l’Education et de la Formation Professionnelle (Ministry for Education and Professional Training) PaP: Port-au-Prince RCT: Randomized Control Trial UFGE: Umbrella Facility for Gender Equality 5 SUMMARY Funded by the World Bank Gender Action Plan (GAP and subsequently the UFGE) and the Nike Foundation, the Haiti AGI is part of the Adolescent Girls Initiative (AGI), an international initiative, to pilot context-specific interventions in low income countries – half of them fragile countries - for a better understanding of the skills that could increase employability for vulnerable young women. Women and particularly young women are underrepresented in Haitian labor markets. They have significantly worse labor indicators than their male counterparts, particularly in labor income, in spite of limited differences in education. For instance, they are 20% more likely to be unemployed than men and, if working, make wages 32% lower. With labor income a key determinant of poverty reduction in Haiti, improving the capacity of young women to access, benefit from, and contribute to the labor market is critical. To that end, the Haiti AGI worked with vulnerable Haitian young women aged 17-21 years in the Port-au-Prince area. The intent was to increase the young women’s agency and capacity to improve their socio-economic situations, particularly through formal labor markets. The overall objective was to improve the employability of the women taking part in this pilot project, i.e. for them to be better positioned to participate productively in the labor market. Between 2012 and 2014, the Haiti AGI provided technical and soft-skills training to over 1,000 young women. The technical skills were in sectors not traditionally associated with female employment but presenting interesting hiring opportunities in a reconstruction/post-reconstruction context: electricity, mechanics, refrigeration, heavy machine operation, telecommunications, and others. Soft-skills sessions were organized in parallel by community organizations working with the beneficiaries in their respective quartiers of residence to boost socio-emotional and non-technical factors such as self-confidence, autonomy, aspirations, and empowerment. The Haiti AGI gathered lessons through qualitative evaluations of the pilot as well as a randomized impact evaluation. Feedback on the quality of the design and implementation from all actors involved was largely positive. A main indicator that the design and implementation were good is that drop-out from the program was extremely low. Of all AGI beneficiaries, only 6% abandoned the program. Typically, technical training programs see drop-out rates in the order of 50% to 60%. The low rate could also be indicative that beneficiaries were satisfied with the project, a finding corroborated by feedback received from the qualitative evaluations conducted in June-July 2013 and December 2014- January 2015. The Haiti AGI has improved the overall agency of beneficiaries, i.e. their capacity to make choices and to pursue desired actions. An assessment three months after program completion found that beneficiaries had higher socio-emotional “assets” compared to women who did not take part in the program. Their autonomy in decision making, their autonomy in personal mobility, and their standing in relations with family and others all rose. Stress declined and mindset became more positive. Beneficiaries’ aspirations about work and income grew more ambitious. The young women were more assertive and talked more to others about job opportunities. Concurring evidence shows increased non-acceptance of violent behavior from personal partners. The integrated design of the 6 project and the importance given to soft-skill training appear to have been key determinants of those outcomes. In terms of labor markets, impacts measured three months after program completion show that beneficiaries were changing the type of work they do, while the share who were participating in income-generating activities, as well as their earnings, had not increased. These short-term impacts show that beneficiaries were transitioning towards higher-skilled jobs related to the technical training they received. Underlying the lower participation in income generation was an enhanced aspiration to pursue further education, necessitating a delay in entering the labor market. Young women who took part in the project displayed higher expectations towards school enrollment and subsequent engagement in the labor force. Encouragingly, these aspirations are materializing in reality with beneficiaries increasing their enrollment for further education. These short-term results show that the Haiti AGI enhanced the employability of young women, improving both technical and socio-emotional skills, all markers of better labor market performance. Right after training, it is not atypical for job-finding rates to be lower for participants than non- participants, as other studies have found. Moreover, the most recent qualitative evaluation that focused on the job search experience of the beneficiaries provided additional keys to understanding the challenges faced by youth and particularly young females in a constrained labor market such as Haiti’s. These include the limited number of job openings, the lack of information regarding those openings, and the common abuse of power (in the form of sexual advances) by men who are in situations of authority in the job application process. The Haiti AGI was a project with an intensive treatment aimed at gaining a more in-depth and contextualized understanding of the constraints and challenges faced by vulnerable youth to participate in the labor market. The Haiti AGI does not aim to be scaled up in its exact current form in the depressed labor market context of Haiti, but could be an option for less constrained environments. Still, elements of the project can be used to inform agency-building and active labor market policies and programs in Haiti and other fragile countries. In light of this, several takeaways are offered for employability-oriented programs and active labor markets programs: 1. Having an integrated project design that builds on local community partnerships and public-private partnerships (PPP) matters for skill-building programs aimed at vulnerable groups. The Haiti AGI confirms the importance of comprehensive intervention programs when dealing with vulnerable groups to address the diversity of challenges they face, beyond gaps in technical skills. Moreover, given the gender equity focus of the project, a takeaway for program designers to consider is that to address gender-specific challenges – including gender-based violence- a network of design elements is advisable, such as limiting commuting time, having alert and monitoring systems, timing activities during daylight. 2. Providing technical training in non-traditional trades is an option for consideration for women’s training programs. Providing training for women in male-dominated fields is not an easy decision for program designers. The takeaway to offer from the Haiti pilot is that participants in non-traditional tracks were as committed to follow-through, remain engaged, 7 and complete the courses and internships as participants who did the traditional tracks in the pilot. While bottlenecks remain to make technical training viable in terms of direct translation into employment, gender does not appear to be one provided the right orientation, consideration to labor market needs and employers’ engagement. 3. Complementing technical training with socio-emotional and soft-skills development can be a cost-effective program components to enhance results. Soft-skills are increasingly recognized for the key role they play in job performance, and as a result are given high importance by recruiting employers. For a fraction of the cost of the technical component, evidence from the project implementation and qualitative evaluations attest to the potential returns of investing in socio-emotional and soft-skills development. More research is needed to further measure those returns. 4. Accounting for aspirations in program design can help channel this dimension towards better outcomes. Aspirations are an important lever for motivation and perseverance. Finding the balance between vulnerable youth aspiration, expectation, and reality directly influences outcomes in terms of program drop-out rates and post-training trajectories. While the Haiti AGI saw an increase in aspirations, the findings also indicate a need for further research on how to best integrate this dimension in future programs. 5. Improving how to provide information and prepare graduates to enter constrained labor markets particularly matters in fragile settings. Educating the trainees about the reality of labor markets both in terms of opportunities and challenges is necessary, but how to do it to maintain motivation and perseverance remains to be further refined. Improving information sharing and transparency so as to increase, and render more inclusive, the matching of labor supply and demand is an important area for further study. 8 I. Introduction The Haiti AGI is part of the worldwide Adolescent Girl Initiative. The AGI was launched on October 10, 2008 as part of the World Bank Group’s Gender Action Plan (GAP), which helps to increase women’s economic opportunities by improving their access to labor market, agricultural land, technology, credit, and infrastructure services. With the financial participation of the Nike Foundation, the AGI piloted context-specific interventions in low income countries – half of fragile countries- for a better understanding of the skills that could increase employability for vulnerable young women. The initiative has been implemented in eight countries: Haiti, Afghanistan, Lao People’s Democratic Republic, Jordan, Liberia, Nepal, Rwanda, and South Sudan. While some on the pilots were eventually scaled up (Liberia, Nepal), the main objective of the AGI has been to inform policies, programs and projects aimed at developing the agency and employability of vulnerable young women. Launched in 2011, the Haiti AGI worked with vulnerable Haitian young women (aged 17-21 years) in the Port-au-Prince area to increase their agency as well as their capacity to improve their socio-economic situations, particularly but not exclusively through formal labor markets. Over the course of the project, more than 1,000 young women were trained, in two cohorts of about 500. Women—particularly young women—are under-represented in Haiti labor markets and have significantly worse labor indicators than their male counterparts (Annex X.3). 1 With labor income a key determinant of poverty reduction in Haiti (see 2014 Haiti Poverty Assessment), improving the capacity of young women to access, benefit from, and contribute to labor markets is critical. The objective of the Haiti AGI has been to increase the employability and the capacity of young women to adapt to Haiti’s complex socio-economic environment. In a constrained setting such as Haiti, any intervention that aims to create full employment would miss the country’s reality, and indeed, many projects that have promised this outcome have failed to meet these expectations. 2 Instead, the Haiti AGI takes a longer perspective. It provides the building blocks for vulnerable young women to not only improve their socio-economic situations in the short term but, more importantly, to develop their capacity to adapt and to find opportunities in a context that is likely to remain highly constrained. 1 Women are significantly disadvantaged in the labor market. Holding constant several social and demographic characteristics, one finds that women are 20 percentage points more likely than men to be unemployed and, if working, 6 percentage points more likely to be in the informal sector. Wages among women are also 32 percent lower than wages among men. Over two-thirds of this difference is unexplained by observable characteristics, suggesting that discrimination could play a role. (World Bank, 2014a). 2 This is the case of the IDEJEN program, a pilot project largely funded by USAID (USAID, 2010). The project has trained a large number of youth (about 11,000) since 2004. It has not been formally evaluated, but employment results have contrasted with anecdotal evidence of overcrowding in some of the trades for which youth were trained in hopes that they would become micro- entrepreneurs. In 2012, IDEJEN became a Haitian NGO. In November 2010, with the support of the IDB, the MasterCard Foundation, the Clinton-Bush Haiti Fund, USAID and others donors, IDEJEN and the NGO YouthBuild International launched the JÈNKA program (the name is a Creole acronym for “Young People Building Haiti”). It has a five-year goal of engaging 9,000 Haitian young people to help rebuild the country after the earthquake. JÈNKA students helped to construct and rehabilitate 12 IDEJEN vocational training centers which host the JÈNKA program and are located in rural and urban places. To date, the total investment in JÈNKA is more than $11 million. The project has not been evaluated. 9 The Haiti AGI is a pilot project whose primary goal has been to identify lessons to improve vulnerable young women’s employability and agency so as to inform larger programs and policies. Its scope reflects this pilot nature and learning focus, to enable intensive support later on to integrate the identified challenges in a multi-component intervention. The Haiti AGI does not aim to be scaled up in its exact current form in the depressed labor market context of Haiti, but could be an option for less constrained environments. Still, elements of the project can be used to inform agency-building and active labor market policies and programs in Haiti. II. Framework of Analysis: Expanded Assets Approach to Supporting Employability The Haiti AGI pilot was designed with the goal of supporting employability. The International Labor Organization (ILO) defines employability skills as the skills, knowledge and competencies that enhance a worker’s ability to secure and retain a job, progress at work, cope with change, secure another job if he/she so wishes or has been laid off, and enter more easily into the labor market at different periods of the life cycle. The AGI focuses on two types of skills to make beneficiaries more employable: technical “hard” skills to improve the competitive profile of candidates in specific areas; and non-cognitive/socio-emotional “soft” skills to strengthen the foundational core aspects of a person’s agency, i.e. self-esteem, self-confidence, aspirations, autonomy. The latter set of skills was identified as particularly relevant for programs working in employability in fragile settings but also reflects the increasing recognition of how cognitive and socio-emotional skills build on each other, starting in early childhood (Helmers and Patnam, 2011). (See also Annex X.4 for a discussion of the concept employability in an expanded framework) The Haiti AGI uses an expanded version of the asset-based framework that links endowment of assets, use (and intensity of use) of assets, and returns to assets, and considers the role of subjective factors such as socio-emotional/soft skills. The assets-based approach can prove useful to study employability, by offering a wider perspective on the drivers and outcomes. Based on the asset-based framework, a person’s income generation capacity depends on Bussolo and Lopez-Calva (2014) and Lopez-Calva and Rodriguez-Castelan (2014):  The endowments of assets;  The use (and the intensity of use) of assets;  The return of assets. Assets are typically understood to include physical and financial assets (private such as tools and bank deposits and public such as roads and welfare support); natural assets (such as land); human capital (education and health); and social capital. In an augmented framework, one can include types of skills that have been increasingly recognized to matter to individuals’ outcomes, namely socio- emotional/soft skills (Duckworth et al., 2007; Heckman et al., 2012; and Cuevas and Favara, 2015). Those skills include key traits and behaviors such as self-confidence, self-efficacy, autonomy of 10 decisions, perseverance, aspirations, and expectations. This socio-emotional capital 3 is also a key aspect of agency, i.e. the capacity to make choices and turn decisions into action, a crucial dimension in the process of achieving development outcomes (Sen 1981, 1999). 4 The capacity to use and convert assets into well-being and desired outcomes differs among individuals and depends on initial endowments and their interaction with the context (markets, institutions, social norms). This interaction defines people’s ability to convert means (e.g. technical and socio-emotional skills) into valuable opportunities (or capabilities, “being able to work”) and outcome (or functioning, “have a job”). 5 Life-skills further complement and enhance those intangible assets, particularly in the case of vulnerable groups. Figure 1. Integrating the asset and the capability approach: The endowment, use and return of assets Source: Cuevas and Favara (2015) based on Bussolo and Lopez-Calva (2014) 3 Socio-emotional aspects are also called psycho-social and soft-skills. The terms are used interchangeably in the report with a preference given to soft-skills. See Annex X.X on soft-skills. 4 Beyond this socio-emotional capital, other authors have also proposed approaches to better account for the subjective dimensions of employability, notably how it relates to not only the way individuals come to perceive and understand the labor market they are entering, but also the types of dispositions, attitudes, and identities they develop around their future work and employability (Tomlinson, 2007). For instance, looking at higher-education graduates, Holmes (2001) develops an “identity approach” to employability, seeing employability as relational, emergent, and influenced largely by one’s “lived experience” of the labor market. Individuals’ experiences of work are subjective, and this is likely to influence their actual labor market outcomes and further shape their propensity for employment. Employability in this sense may be seen to be value- and identity-driven, relating to graduates’ own pre-dispositions and biographies. 5 The framework helps explain why individuals endowed with the same assets are not equally successful in the process of making choices, converting those choices into actions and finally achieving the desired outcome of converting those assets into livelihood (Sen, 1981 and 1999). In so doing, the proposed framework also enables the consideration of dimensions highlighted in the 2014 World Development Report (Mind, Society and Behavior) that recognized the role of social expectation in sharing individuals’ preferences. These preferences are themselves dependent on the context in which they are elicited and on the social institutions that have formed the interpretative frameworks through which individuals see the world (Basu, 2010; Fehr and Hoff, 2011; World Bank 2014a). Endowment of Assets •Tangible assets (physical assets, financial assets, natural capital) •Intangible assets (human capital; social capital; socio-emotional capital) Use of Assets •Ability to take decisions and use the assets. Return of Assets •Conversion into outcomes Contextual enablers Institutional enablers 11 The assets-based conceptual framework has been supported by academic research and has also been extensively used in other studies that have analyzed the determinants of progress of poverty reduction and shared prosperity around the world (for these issues, see, for instance, Attanasio and Székely, 1999 and Carter and Barrett, 2006). This model is particularly useful to present a characterization of intra- and inter-generational economic mobility, chronic and transient poverty, and between-group inequities (poor and non-poor, bottom 40 and top 60, minorities, etc.) that potentially thwart certain vulnerable populations from fully participating in and benefiting from the development process. Furthermore, it is useful to consider agency and its multiple dimensions at two levels: the individual in relation to others, and the individual in relation to self. Agency is the capacity to make choices and translate them into desired actions and outcomes. Because agency is a complex and encompassing construct, it is useful to break down its analysis by distinguishing aspects of the individual in relation to others and the individual in relation to self. For the former, this report focuses on autonomy of decision making, autonomy of mobility, and quality of relationships. For the latter, the report analyzes includes self-esteem, confidence, capacity to aspire, mindset, and stress. Figure 2 summarizes these elements. Figure 2. Breaking down agency and socio-emotional capital In particular, aspirations (and capacity to aspire) is growingly acknowledged as an important element for development outcomes. The role of aspirations in development has gained increased recognition in the economic literature over the past decade (see Box 1). Leading researchers such as Barnejee and Dufflo (2012), Dercon and Singh (2013), and Ray (2006, 2010) have helped give more visibility to the role of aspirations, notably in the perpetuation of “poverty traps” in which people are trapped in a self-reinforcing cycle of poverty, low aspirations, and sense •Autonomy of decision making: labor, education, spending decisions •Autonomy of mobility: able to attend places that the individual values •Relationships: having a strong support network, being respected by domestic partner Individual in relation to others •Self-esteem, self-confidence to pursue desired actions •Stress, anxiety, depression as mental statuses that curtail brain capacity for decision making and action •Mindset: state of mind as a resource to take right decisions, persevere, and achieve desired outcomes Individual in relation to self Agency and socio-emotional capital 12 of helplessness. 6 The World Bank has come to recognize aspirations’ importance in the reduction of poverty and the expansion of prosperity, notably through the promotion of empowerment and agency (Alsop et al., 2006; World Bank, 2012; and World Bank 2014b). The concept of aspiration has direct application to employability through “aspirations traps.” Alsop et al. (2006) documents how sense of inferiority helps cause underinvestment and subsequent under-accumulation of human capital by women and minority groups: “[Such] groups frequently underinvest in their human capital because they have been brought up to believe that they cannot do certain things that other people can do. [They] internalize their second class status in ways that cause them to make choices that perpetuate their disempowered status.” Box 1 : The role of aspiration in the reduction of poverty and the expansion of shared prosperity The role of aspiration in development has gained wider recognition over the past decade, although its origins can be traced to older literature. In his seminal book La Distinction (1979), Pierre Bourdieu describes how a person’s social environment generally determines his or her interests, tastes, and ideas about life. During the process of socialization, children learn what “children like them” typically believe and enjoy doing (Bourdieu called this habitus). The children then incorporate these social norms into their own lives. In this way, the subconscious incorporation of norms, aspirations, and goals influences careers and other life choices. People usually consider pursuing only the options with which they are familiar. For example, if going to college is not usually an option for “a person like you,” you will not consider it. This approach emphasizes how aspiration levels and living standards are transmitted from one generation to the next. More recent literature has demonstrated the relevance of those findings and their implication for development. For instance, Dercon and Singh (2013) demonstrate these transmissions with data from Ethiopia and India. In these cases, parents tend to have higher aspirations for boys than for girls. Over time, as the children became older, they too assimilate the respective aspirations of their parents. The study found that, eventually, the educational levels of boys and girls reflect these aspirations, with boys attaining higher levels of education. Bourdieu’s work also suggests an important link between aspirations and identity. When people construct an opinion about what is possible for “people like them,” they consider the experiences and aspirations of peers (Ray, 2006). As a result, role models can have an important impact on people’s aspirations in both a positive and negative way. For instance, Beaman et al. (2009) showed that the presence of women in leadership positions in India lessened the gender bias in parents’ aspirations. In Madagascar, Nguyen (2008) investigated the effect of role models on aspiration levels. The author carried out an experiment in which people with successful educations shared their experiences with skeptical parents in order to improve how much value the parents would place on education. The study found that parents tended to be influenced only if the “role model” came from a similar socioeconomic background—in this case, from a poor background. 6 On the other hand, more privileged individuals tend to be more optimistic and even upwardly biased about their capacities. As noted by Wilkinson and Pickett (2010), “the further up the social ladder you are, the more help the world seems to give you in keeping the self-doubts at bay.” Stutzer (2004) confirms that income aspirations grow with higher income levels. 13 Source: Adapted from Flechtner (2014) Making use of an augmented assets approach, the study proceeds to analyze whether a training program that increases the endowments of human capital and socio-emotional assets of its beneficiaries is able to improve their employability. III. The Haiti AGI project Informed by extensive consultations, the integrated design of the Haiti AGI placed the beneficiaries at the heart of a network including community NGOs, training centers, and employers. Building on the identified issues and challenges to the agency and employability of young females, the Haiti AGI team and partners set out to develop a project comprehensively integrating different components aimed at addressing the facets of the vulnerability experienced by these young women. Figure 3: The Haiti AGI integrated project design The Haiti AGI prioritized working with recognized community-based organizations to ensure proximity to and familiarity with the communities where the beneficiaries reside and to build the capacity of Haitian NGOs. The five community NGOs selected for the project are located in five quartiers (neighborhoods) of Port-au-Prince. Following a public expression of interest (EOI) and consultations with the Gender Ministry (MCFDF) and other partners (international organizations and NGOs), the community NGOs were selected. They were picked on the basis of their experience working with vulnerable young women in the five quartiers and their management capacity to provide accompaniment services and deliver soft-skill training to an average of 100 young women in each of two cohorts for two periods of six months. 7 Five NGOs were eventually 7 Community NGOs with a lesser capacity, for instance due to a smaller building capacity to host the soft-skills sessions, were asked to follow a smaller number of young women (no less than 75), while those with more capacity hosted a larger cohort of up to 150. 14 identified. 8 Close monitoring and capacity-building training were provided to ensure the effective management of the project. Box 2. Selection process of the Haiti AGI beneficiaries Targeting. The target population of the project is out-of-school vulnerable young women aged 17-21 in five poor neighborhoods (quartiers) of Port-au-Prince, Haiti. The neighborhoods—Delmas 30, Delmas 32, Carrefour-Feuilles, Martissant, and Pétionville—were selected based on the poorer socio-economic profiles of their inhabitants (see map of Port-au-Prince and the locations of the quartiers in Annex X.1.) The selection of the training centers and community NGOs also factored in the home locations of the beneficiaries to limit the commute, ensure the proximity of the accompaniment of the beneficiaries in their community, and limit safety risks. The project was funded to provide training to a total of 1,000 young women. Selection criteria: Consultations with NGOs (local and international), training centers, social and education specialists as well as employers were used to refine the criteria used to select the project’s participants. A review of comparable programs and projects (including other AGI pilots) helped confirm those criteria and their relevance for the development objective of the project. Five main criteria were defined: - Female gender (given the AGI focus on young women) - Age: 17-21 years - Out of school for over a year - Basic literacy and numeracy ability (as demonstrated by passing an academic test) - Living in one of the five selected neighborhoods of Port-au-Prince. In the consultations, enforcement of a household or per capita income threshold was not deemed feasible. Instead, the selection relied on the quartier of residence of the young women and the knowledge of the community NGOs of the background of the candidates. Recruitment. The community NGOs ran campaigns and mobilization efforts to recruit young women for the project. Many self-identified, which demonstrated a pre-existing level of motivation and interest, and is considered to be important for the objective of the program. The young women were required to take an academic test to ensure basic levels of literacy and numeracy. The World Bank worked closely with the firm hired to collect data for the impact evaluation to make sure that all young women selected were of the right age and level of academic ability. Academic testing. Imaginescence, a Haitian company that specializes in educational testing, was hired to prepare and administer the exam. This testing of numeracy and literacy skills (reading, writing, and math skills) was necessary to ensure that the candidates had a sufficient academic level to enable them to follow the proposed technical training. In spite of guarantees provided by the firm, the test proved to be too difficult for and not well adapted to the academic level of the young women. 9 The test scores were very low. Still, the test provided a baseline and a metric for qualification. Subsequent “entry testing” by the training firms confirmed that the young women had the required levels of numeracy and literacy. 8 The community NGOs are: APROSIFA (Carrefour-Feuille), ANAPFEH (Delmas 32), YWCA (Pétionville), COFEHAPS (Martissant), and JP/HRO (Delmas 30). While JP/HRO qualifies as an international organization, the organization’s mandate is focused on Haiti and the organization has been well-established in community work since the earthquake. 9 The academic format of the test appears to have intimidated a number of the candidates. 15 Cohort assignment. Due to the design of the project and its impact evaluation, the beneficiaries were divided into two cohorts through a lottery-based randomization process conducted in each of the quartiers in October 2012 under the supervision of the project field coordinator. “Non-traditional” technical training was selected for the project based on extensive consultation with stakeholders to seize on existing employment opportunities. Informed by in-depth consultations with employers and other stakeholders, the project focus was set on non- traditional technical training (i.e. training not typically associated with female employability interventions) so as to seize upon opportunities linked to post-earthquake reconstruction both in the formal and informal sector. The employer assessment conducted in 2011 also identified interest by employers to hire more women, who were considered to display higher work ethics than men. Converging evidence on the evolution of social norms regarding acceptable sector and jobs for women also presented an opportunity for the AGI pilot to better understand the types of skills and support needed to overcome gender-specific barriers to employability. The technical training was provided by public and private centers selected notably for the recognized quality of their training and their labor-market orientation. Extensive consultations, notably with employers, and an assessment by the American Institutes of Research were conducted in the design phase of the project regarding the availability of technical training in Haiti. 10 This served to evaluate proposals received through the EOI in early 2012. 11 In a sector dominated by private providers, the bids varied greatly in quality, relevance, and price. Eventually four private training providers were selected to complement the public offer by INFP. 12 While a mixed approach of public and private provision of technical skills was envisioned from the start, this PPP approach was further mandated by the limited training capacity of the INFP following the 2010 earthquake, which had destroyed all of its centers. 13 One of the training centers (Centre Polyvalent de Formation Professionnelle de Carrefour— CPFPC) was under reconstruction at the time of the first cohort’s training but was integrated as a training provider for the second cohort. Still, the INFP’s involvement in the project was critical to ensuring that the lessons learnt through the project would be welcomed and possibly adopted by the relevant ministerial counterparts. The selection of private providers certified by the INFP also 10 In 2011, American Institutes for Research (AIR)—a not-for profit organization specializing in behavioral, social research, and evaluation—was contracted to undertake an assessment of the available supply of vocational training centers in Haiti., with emphasis on management and operation and links to labor markets. The final report (“Promoting the Economic Empowerment of Adolescent Girls in Haiti: Definitions, Design Options and Project Profiles”) was submitted to the World Bank in December 2012. 11The selection criteria for the training providers included: A solid reputation among employers in Port-au-Prince; at least five years of experience successfully implementing vocational training programs, with those targeted to young women considered a plus; solid technical training materials that have been successfully tested and meet criteria for INFP’s certification; evidence-supported connections to the private sector; a job placement office or staff allocated for this purpose ; demonstrated success placing graduates in jobs; experienced and well-respected teachers; and neighborhood location of the training center which must also be a space where participants feel safe and comfortable. 12 The four training centers selected were: APEX, Haiti Tec, INFOP, and Transversal. 13 Effectiveness of private-sector providers of technical trainings has been better documented in developed countries, where non- experimental literature has found evidence that the impacts of training is higher when provided by private providers (Jespersen et al., 2008). Provided that the selection of training providers is adequate, possible reasons are that private providers are more responsive to private-sector employer demand and/or they potentially face more competition and thus must increase quality in response. A recent experimental evaluation of active labor market programs in Turkey shows that contracting with private providers can substantially improve results (Hirshleifer et al., 2014). 16 helped reinforce the links between private and public training providers, further deepened through the regular project monitoring meetings organized at the World Bank office with representatives of the respective training providers. Training to develop entrepreneurship and self-employment skills was considered, but ultimately it was decided to focus on technical training. 14 While entrepreneurship can be an important option for vulnerable youth programs, such programs have wide variation in effectiveness, and do not necessarily translate into faster business setup and expansion or increased income (Cho and Honorati, 2013). It is challenging to identify beneficiaries that are predisposed to become successful entrepreneurs in the context of Haiti. This is why the project design rather focused on building the human capital of the young women through training and on-the-job experience in trades also potentially offering self-employment opportunities. Box 3: Building partnerships with ministries While the Haiti AGI was implemented by the World Bank, as a result of the 2010 earthquake, it strived to develop a strong client approach from the start. The Ministry of Education and Technical Training (Ministère de l’Education et de la Formation Professionnelle- MENFP)-, and the Ministry of Gender Affairs (Ministère de la Condition Féminine et aux Droits des Femmes - MCFDF) were consulted early on to develop the design and the gender focus of the AGI. During implementation, a partnership was developed with the Technical Training Secretariat (INFP, within the Ministry of Education), which is ideally positioned to further incorporate lessons into ongoing and future programs. Issues such as soft-skills and M&E were identified to the counterparts as being of particular interest and were therefore received priority in discussions and technical workshops. In a context of high ministerial turnover and political instability, the project team was able to build trust and support for the project as a platform for cooperation. Moving forward, this strong client focus in a context of highly constrained client capacity, combined with the solid civil society partnerships and labor market orientation of the project, also provided the Bank with unique direct experience and made it a recognized voice on the issues related to skills/technical training, youth and jobs. These were all identified priorities of the government’s three-year investment plan (Plan Triennal d’Investissement, 2014-2016), as well as the recently completed Haiti Systematic Country Diagnostic (World Bank, 2015). Employers have been involved at all stages of the Haiti AGI. Maintaining connection with the private sector and providing technical training reflecting short- to medium-term opportunities in Haiti’s labor market have been key priorities of the Haiti AGI. While informality dominates Haiti’s labor markets, the Haiti AGI opted to provide quality technical training that would allow for graduates to move into jobs in the formal sectors. All the while it recognized the reality that informal sectors and self-employment would likely be the receiving segments of Haiti’s labor markets for the AGI trainees. 15 Employers from the sectors for which the AGI beneficiaries were being trained were 14 A partnership with IFC to provide the “Business edge training” to 100 young women displaying the most potential in terms of entrepreneurship qualities (to be identified by the training centers) was envisioned but could not be implemented for funding reasons. 15 Results from the 2012 ECVMAS show that outside of the primary sector, self-employment and informality are the rule: 44% of the workforce is employed at the informal private sector. That means that close to 90% of workers in Haiti are outside the modern formal 17 directly connected to the project through regular consultations, and by the project team and training centers’ preparation for internships. 16 These one-month internships provided the young graduates with direct experience in their sectors. 17 Internships are an important employment placement instrument, but they remain unusual in Haiti. The Haiti AGI placed a premium on providing soft-skills training. Building on best practices worldwide, the Haiti AGI developed a high-quality soft-skill and life-skills curriculum of eight modules: Self-esteem; civic engagement and leadership; reproductive health; gender-based violence and responses to it; preparing for the workplace and work ethics; disaster preparedness; financial literacy; and living with a disability. 18 The mix of soft-skills and life-skills was informed by the focus of the project on vulnerable young women and the specific context of Haiti. 19 Existing curriculum proved to be ill-adapted to the Haitian context, being mere translation of curriculum developed for other countries, notably in Africa. Based on consultations with partners, the Haiti AGI worked with Haitian education specialists and illustrators to develop a curriculum adapted to the context of Haiti. The soft-skills training was administered through face-to-face student-centered sessions taking place on week-ends on a morning/afternoon rotating basis, at the office of the respective community NGOs and in parallel with the technical training. Based on consultations with the community NGOs to define a standardized method to teach this curriculum, the training prioritized interactivity and role-playing. Stipends to cover the costs of attending training were provided to the AGI beneficiaries, using cell-phone based technology (mobile money) to reduce risks and increase control over the funds. The Haiti AGI provided stipends to enable the young women to cover minor costs associated with their participation in training. The stipend aimed to cover the cost of transportation to and from the training centers and other related expenses for attendance. The stipends amounted to about US$40 per month, a level defined based on consultation with community NGOs and on training centers’ experience working with other international partners that had provided stipends to trainees. The overall financial capacity of the project was also a consideration. 20 The delivery of the sector, be it private, public or in NGOs. Among workers in the informal sector, almost 60% work as self-employed or unpaid workers in family firms. Wage employment is much more prevalent at the Metropolitan area of Port-au-Prince (32%) and other urban areas (27%) than in the countryside (16%). The vast majority of workers in the informal sector, furthermore, are employed in low- productivity sectors—six out of ten work in the traditionally low-productivity retail and wholesale sector. Another 15% are employed in other services and 8% in construction. 16 The project team had envisioned hosting monthly or bi-monthly breakfast workshops with selected employers to discuss their management and human resource experience, but this did not turn out to be feasible, for reasons of cost as well as limited capacity to dedicate to organizing those events. In addition, feedback suggested that attendance would be limited. 17 The one-month internships were unpaid but the young women received their stipends during them. 18 This eighth module was developed and financed through a grant provided by the Y2Y fund. The recipient of the grant, Francesca de Antoni, worked with the Haiti AGI team and other partners of the AGI to develop the corresponding module. 19 The terminology of soft-skills has been used to characterize skills that have also been defined as socio-emotional in recent and still evolving literature on the issue. The authors recognize the typology of skills emerging from initiative such as the World Bank Skill Measurement Program (STEP) and its corresponding surveys, but elected the concept of ‘soft-skills’ as the AGI intervention did not target all the skills associated with the terminology defined by the STEP survey. 20 The amount was considered to be on the lower end of typical stipends provided at the time in comparable projects in Haiti (more in the range of US$80-$100), with some concern that the beneficiaries might consider the amount too low. But they did not express such misgivings and no drop-out was linked to the stipend level. In fact, in a private center that simultaneously hosted the AGI trainees and a group of youth sponsored by another international organization, center officials noted the higher participation rate and motivation of the AGI trainees. 18 stipend via a mobile phone payment system was selected in order to increase safety and avoid risks of handing over cash to the women and to help enforce the financial management lessons provided during training. 21 The Haiti AGI stressed gender-informed safety and security throughout the project cycle and design. As a project focused on vulnerable young women, the Haiti AGI placed a premium on the physical safety of its participants. Port-au-Prince is well known for its volatility and protracted insecurity, a concern heightened in the quartiers of the AGI project (notably Carrefour-Feuille and Martissant). While this reality applies to all inhabitants of Port-au-Prince, the poor and vulnerable are more directly exposed to it because their neighborhoods are often the focus of gang-led violence, and they have fewer options they have to protect themselves. One of the greatest risks facing the Haiti AGI beneficiaries was physical assault, particularly sexual violence. Such violence can be experienced in the private as in the public sphere, so the project implemented different approaches to address it:  Limiting commuting time: The project design recognized the importance of limiting the commuting of the AGI trainees to reduce their exposure to assault when traveling to or from their courses. Beneficiaries were matched to training centers located within the most reasonable commuting time feasible, on average 30 minutes. Technical training offering employment opportunities in specific locations of Port-au-Prince (for example, the SONAPI industrial park) were flagged as such to the trainees. Those residing in more distant quartiers were asked to consider the time and money costs resulting from the needed commute before finalizing their technical training choice. This approach was also applied to the internship component of the project: the trainees were matched to internships within reasonable distance from their quartier of residence.  Training schedule: The training schedules proposed half-day classes taking place in the morning, allowing the young women to travel to and from during daylight.  Gender-sensitive training: Gender-sensitive training led by the field project coordinators was organized in the different technical training centers to sensitize trainers to working with vulnerable young women. The training aimed at preventing abuse and inappropriate situations, as well as identifying behaviors (e.g. restlessness, acting-up, depression, and absence) by the women that might indicate a situation of abuse or assault.  Mentors as sentinels: Due to their proximity to the beneficiaries, the mentors played a key role in sensitizing about sexual violence, providing information regarding prevention and recourse in case of sexual violence, and accompanying young women to request services or assistance related to violence. 22 21 The first installments were paid through the training center partners due to delays in the setting up of the Tcho-Tcho system of the cellphone operator Digicel. 22 The AGI team facilitated the emergency housing of several AGI beneficiaries facing situations of domestic violence, in partnership with the community NGOs. 19  Using cell phones for communication and stipend payment: The beneficiaries were provided with cell phones and some minimal air time on the Digicel network to enable them to give the alert and request assistance in case of dangerous situations. 23 The cell phones were also used to distribute the stipends to the beneficiaries so as to limit their risk of being robbed and to provide them with a greater control over the use of those funds. In a country where gender-based violence is “committed with near complete impunity” (OHCHR, 2012), a network of measures and interventions is needed to prevent and address it. Mechanisms to prevent and respond to violence against women remain limited and the justice system ill-equipped to deal with this pressing issue (MINUSTAH, 2013). 24 In light of this bleak reality, the Haiti AGI opted to place a strong focus on prevention of this violence and accompaniment of young women facing it, notably through the network of NGOs providing support that is more targeted to this type of violence and care. Monitoring and evaluation has been central to the Haiti AGI. As a pilot, the Haiti AGI had evaluation as a key dimension. On the monitoring side, an online platform was developed to facilitate the tracking of information regarding the beneficiaries’ technical and soft-skill training (courses followed, grades, and attendance). During the first cohort, local partners received extensive capacity building in how to monitor the program through the online platform. However, in spite of this training and the support of the M&E field coordinator, the platform turned out to be too difficult for some partners to use. Given the importance of providing timely information on the trainees to enable partners to act—for instance, with respect to absence or dropping grades—a more user-friendly Excel-based system was used for the second cohort. This system, based on weekly submission of grades and attendance by the training centers and community organizations, required the M&E coordinator to manually aggregate the results and send back a summary report to the partners for action. This approach was rendered possible by the capacity-building investment during the first cohort and the familiarity of the partners with the reporting process. This system was instrumental to monitor and manage program drop-out. (See. Table 1) Table 1. Graduation and abandonment rate by cohort – Haiti AGI AGI trainees Graduated Drop-out/abandonment 1st cohort Number 473 433 40 Percentage 100.0 91.5 8.5 2nd cohort Number 536 517 19 Percentage 100.0 96.5 3.5 Total Number 1009 950 59 23 The young women were provided with emergency numbers (police, NGOs specializing in sexual violence such as KOFAVIV, mentors numbers and project field coordinator number). 24 Concerning rape, while the Haiti penal code has recognized rape since 2005, it does not define the acts constituting such an offense. Discussions in 2012 related to a wide revision of the Haitian Penal Code included a definition of rape, but so far no changes have resulted. Furthermore, many challenges, including the victim’s need to produce a medical certificate, the push by judges for cases to be resolved through “mediation and reparation,” and the overall lack of rigor and neglect in the registering of cases add to the overall complexity of seeing victims heard and perpetrators brought to justice. (MINUSTAH, 2013). [... middle sections omitted for long document ...] 120 Table X12. Impact of the AGI on the level of satisfaction with current status Baseline Mean OLS OLS OLS Obs. How satisfied are you with: 1. Your education level? [score: 1-4] 2.623 -0.219 *** -0.220 *** -0.217 *** 2326 [ 0.808] ( 0.065) ( 0.062) ( 0.054) 2. Your relationship with your family? [score: 1-4] 2.087 -0.049 -0.049 -0.045 2324 [ 0.627] ( 0.051) ( 0.051) ( 0.044) 3. Your relationship with your friends? [score: 1-4] 2.221 -0.043 -0.043 -0.044 2320 [ 0.611] ( 0.052) ( 0.052) ( 0.048) 4. Your relationship with your husband/boyfriend? [score: 1-4] 2.570 -0.100 -0.103 -0.118 2041 [ 0.938] ( 0.085) ( 0.085) ( 0.077) 5. Your job or business? [score: 1-4] 3.574 -0.038 -0.063 0.037 1642 [ 0.712] ( 0.073) ( 0.069) ( 0.080) 6. Your earnings/income? [score: 1-4] 3.569 -0.075 -0.091 -0.077 1717 [ 0.689] ( 0.068) ( 0.065) ( 0.075) 7. The house you live in? [score: 1-4] 2.604 0.048 0.048 0.059 2321 [ 0.747] ( 0.063) ( 0.062) ( 0.054) 8. The community you live in? [score: 1-4] 2.768 0.026 0.025 0.033 2324 [ 0.747] ( 0.062) ( 0.061) ( 0.052) 9. Life as a whole? [score: 1-4] 2.510 -0.067 -0.068 -0.072 2321 [ 0.739] ( 0.059) ( 0.058) ( 0.055) Satisfaction (sum) [score: 1-36] 21.352 -0.444 -0.444 -0.444 2348 [ 5.830] ( 0.460) ( 0.459) ( 0.432) Satisfaction PCA score (standardized) -0.000 -0.059 -0.061 0.011 1597 [ 1.000] ( 0.108) ( 0.106) ( 0.098) Controls No Yes No Using individual fixed effects No No Yes For all tables: Standard deviation is in brackets. Robust standard error is in parentheses. Level of Significance: * p<0.1; ** p<0.05; *** p<0.01. Notes: The variables used as controls in the regressions are engagement in income-generating activities, dummies for groups of age and education of respondents, household size and the number of female members in the household, school assistance, an indicator of whether the individual has children and marital status of the respondent. All controls are from baseline data. The sum of points of the satisfaction variable is the total score of all the satisfaction statements combined, where the points given to each statement are negatively related to satisfaction: 1 for "Strongly Agree," 2 for "Agree," 3 for "Disagree," and 4 for "Strongly disagree." This means that the higher the total score, the lower satisfaction the respondent has. PCA stands for principal component analysis, the method used to consider the score of all statements and formulate them in a standardized index. 121 Table X13. Impact of the AGI on the level of mental health Baseline Mean OLS OLS OLS Obs. During the past 15 days how often did you: 1. Feel depressed? [score: 1-4] 2.286 -0.011 -0.011 -0.015 2319 [ 0.959] ( 0.077) ( 0.077) ( 0.072) 2. Feel overwhelmed? [score: 1-4] 2.584 -0.177 ** -0.176 ** -0.175 ** 2326 [ 1.061] ( 0.086) ( 0.086) ( 0.080) 3. Feel crying or tearful? [score: 1-4] 2.374 -0.060 -0.060 -0.060 2325 [ 1.006] ( 0.084) ( 0.084) ( 0.078) 4. Feel sad, helpless in life and without hope? [score: 1-4] 2.178 -0.122 -0.121 -0.107 2325 [ 1.050] ( 0.087) ( 0.086) ( 0.082) 5. Have trouble falling asleep? [score: 1-4] 2.096 -0.032 -0.033 -0.023 2324 [ 1.052] ( 0.087) ( 0.086) ( 0.081) 6. Feel tired or missing force? [score: 1-4] 2.306 0.002 0.003 0.003 2326 [ 0.995] ( 0.082) ( 0.081) ( 0.080) 7. Have no appetite? [score: 1-4] 2.016 -0.087 -0.086 -0.072 2322 [ 1.088] ( 0.091) ( 0.090) ( 0.093) 8. Feel your life went wrong or you feel you are a failure? [score: 1- 4] 1.748 -0.146 * -0.144 * -0.147 * 2323 [ 1.027] ( 0.080) ( 0.078) ( 0.075) 9. Want to be dead or harm yourself? [score: 1-4] 1.433 -0.098 -0.096 -0.103 2322 [ 0.841] ( 0.068) ( 0.067) ( 0.063) 10. Not feel your hand? [score: 1-4] 1.809 -0.168 ** -0.168 ** -0.172 ** 2321 [ 0.902] ( 0.075) ( 0.075) ( 0.075) 11. Feel that your body felt warm, but not because of the weather or because you were sick? [score: 1-4] 1.744 -0.074 -0.073 -0.073 2322 [ 0.897] ( 0.073) ( 0.073) ( 0.068) 12. Feel your feet/hands trembling? [score:1-4] 1.752 -0.091 -0.091 -0.077 2323 [ 0.912] ( 0.073) ( 0.073) ( 0.069) 13. Feel dizzy? [score:1-4] 1.997 -0.158 ** -0.156 ** -0.146 ** 2323 [ 0.966] ( 0.078) ( 0.078) ( 0.070) 14. Have palpitations? [score:1-4] 1.527 -0.094 -0.093 -0.096 2323 [ 0.815] ( 0.069) ( 0.069) ( 0.061) 15. Have trouble breathing? [score:1-4] 1.398 0.086 0.087 0.086 2319 [ 0.730] ( 0.062) ( 0.062) ( 0.058) 16. Feel intimidated? [score: 1-4] 1.626 -0.014 -0.014 -0.019 2324 [ 0.858] ( 0.069) ( 0.068) ( 0.067) 17. Have indigestion? [score: 1-4] 1.399 -0.098 * -0.097 * -0.105 ** 2321 [ 0.690] ( 0.055) ( 0.055) ( 0.052) 18. Feel like you will faint? [score: 1-4] 1.400 -0.068 -0.067 -0.060 2318 [ 0.712] ( 0.058) ( 0.058) ( 0.052) Mental health (sum) [0-72] 33.208 -1.271 * -1.270 * -1.271 * 2348 [ 9.362] ( 0.764) ( 0.755) ( 0.685) Mental health PCA score (standardized) -0.000 -0.175 ** -0.172 ** -0.169 ** 2295 [ 1.000] ( 0.089) ( 0.088) ( 0.079) Controls No Yes No Using individual fixed effects No No Yes For all tables: Standard deviation is in brackets. Robust standard error is in parentheses. Level of Significance: * p<0.1; ** p<0.05; *** p<0.01. Notes: The variables used as controls in the regressions are engagement in income-generating activities, dummies for groups of age and education of respondents, household size and the number of female members in the household, school assistance, an indicator of whether the individual has children and marital status of the respondent. All controls are from baseline data. The sum of points of the mental health variable is the total score of all the satisfaction statements combined, where the points given to each statement are negatively related to mental health: 1 for "Not at all," 2 for "1-5 days," 3 for "6-9 days," and 4 for "Almost every day." This means that the higher the total score, the lower mental health the respondent has. PCA stands for principal component analysis, the method used to consider the score of all statements and formulate them in a standardized index. 122 Table X14. Impact of the AGI on attitudes towards violence Baseline Mean OLS OLS OLS Obs. Physical violence with your husband/boyfriend [yes=1] 0.059 -0.025 -0.022 -0.046 1288 [ 0.236] ( 0.028) ( 0.028) ( 0.035) Is it acceptable for a husband or boyfriend to hit his wife or girlfriend if she: 1. Argues with him and insults him [score: 1-4] 1.038 -0.016 -0.016 -0.016 2326 [ 0.270] ( 0.022) ( 0.022) ( 0.022) 2. Goes out without telling him [score: 1-4] 1.086 -0.018 -0.018 -0.020 2324 [ 0.357] ( 0.032) ( 0.032) ( 0.033) 3. Neglects the children [score: 1-4] 1.082 -0.049 * -0.049 * -0.043 2324 [ 0.378] ( 0.029) ( 0.028) ( 0.028) 4. Neglects the household duties [score: 1-4] 1.087 -0.024 -0.023 -0.017 2321 [ 0.404] ( 0.032) ( 0.032) ( 0.032) 5. Questions him about cheating [score: 1-4] 1.651 -0.206 ** -0.206 ** -0.205 ** 2322 [ 0.999] ( 0.083) ( 0.083) ( 0.083) 6. Cheats with another man [score: 1-4] 1.247 -0.158 *** -0.158 *** -0.153 *** 2321 [ 0.688] ( 0.050) ( 0.049) ( 0.048) 7. Refuses to have sex with him [score: 1-4] 1.370 -0.121 ** -0.121 ** -0.119 ** 2325 [ 0.722] ( 0.059) ( 0.060) ( 0.060) Violence towards partner (sum) [score: 1-28] 8.536 -0.614 *** -0.615 *** -0.598 *** 2327 [ 2.276] ( 0.185) ( 0.184) ( 0.181) Violent attitude towards partner PCA score (standardized) 0.000 -0.182 * -0.182 * -0.175 * 2317 [ 1.000] ( 0.096) ( 0.094) ( 0.093) Controls No Yes No Using individual fixed effects No No Yes For all tables: Standard deviation is in brackets. Robust standard error is in parentheses. Level of Significance: * p<0.1; ** p<0.05; *** p<0.01. Notes: The variables used as controls in the regressions are engagement in income-generating activities, dummies for groups of age and education of respondents, household size and the number of female members in the household, school assistance, an indicator of whether the individual has children and marital status of the respondent., All controls are from baseline data. The sum of points of the violence variable is the total score of all the violence statements combined, where the points given to each statement are positively related to violence: 1 for "Never," 2 for “Sometimes,” 3 for "Usually," and 4 for "Always." This means that the higher the total score, the higher attitudes toward violence the respondent has. PCA stands for principal component analysis, the method used to consider the score of all statements and formulate them in a standardized index. 123 Table X15. Impact of the AGI on behavior Baseline Mean OLS OLS OLS Obs. Are you pregnant? [yes=1] 0.014 0.007 0.007 0.006 2270 [ 0.119] ( 0.013) ( 0.013) ( 0.013) Are you planning to be pregnant in the next two years? [yes=1] 0.030 0.031 * 0.031 * 0.029 ** 2302 [ 0.170] ( 0.016) ( 0.016) ( 0.015) Always uses a condom [yes=1] 0.523 0.055 0.052 0.058 1554 [ 0.500] ( 0.053) ( 0.052) ( 0.053) Have you ever tried cigarettes? [yes=1] 0.028 -0.020 -0.020 -0.021 * 2323 [ 0.166] ( 0.013) ( 0.013) ( 0.012) Have you ever tried recreational drugs? [yes=1] 0.002 -0.001 -0.001 -0.001 2313 [ 0.042] ( 0.003) ( 0.003) ( 0.003) Drinks alcohol often (more than 3 to 5 times a week) [yes=1] 0.046 0.027 0.026 0.030 2275 [ 0.210] ( 0.018) ( 0.018) ( 0.019) Alcohol drank last month (Categ: 1-7) 1 5.324 -0.632 *** -0.595 *** -0.364 1027 [ 1.636] ( 0.214) ( 0.214) ( 0.280) Controls No Yes No Using individual fixed effects No No Yes For all tables: Standard deviation is in brackets. Robust standard error is in parentheses. Level of significance: * p<0.1; ** p<0.05; *** p<0.01. Notes: The variables used as controls in the regressions are engagement in income-generating activities, dummies for groups of age and education of respondents, household size and the number of female members in the household, school assistance, an indicator of whether the individual has children and marital status of the respondent,. All controls are from baseline data. 1 The seven categories of this question are 1 for "Every day or almost every day," 2 for "3 to 5 times a week," 3 for "1 or 2 days a week," 4 for "2 or 3 days a month,",5 for "Once a month or less,", 6 for "1 or 2 days in the past 12 months," and 7 for "Never." 124 Table X16. Impact of the AGI on networks and spillover on family members Baseline mean OLS OLS OLS Obs. You go to no one in order to talk about plans for the future [yes=1] 0.130 0.087 *** 0.087 *** 0.087 *** 2348 [ 0.337] ( 0.030) ( 0.030) ( 0.029) You go to friends in order to talk about work problems [yes=1] 0.044 0.015 0.015 0.015 2348 [ 0.206] ( 0.019) ( 0.019) ( 0.019) You go to husband/boyfriend in order to talk about money problems [yes=1] 0.169 0.047 0.047 0.047 * 2348 [ 0.375] ( 0.033) ( 0.032) ( 0.027) You go to husband/boyfriend in order to talk about health problems [yes=1] 0.120 0.045 0.046 0.045 * 2348 [ 0.325] ( 0.029) ( 0.028) ( 0.024) You go to husband/boyfriend in order to talk about problems with your children [yes=1] 0.043 0.028 0.028 * 0.028 * 2348 [ 0.204] ( 0.019) ( 0.017) ( 0.016) You go to mother in order to talk about work problems [yes=1] 0.117 0.037 0.037 0.037 2348 [ 0.321] ( 0.029) ( 0.029) ( 0.029) You go to mother in order to talk about family problems [yes=1] 0.325 0.044 0.044 0.044 2348 [ 0.468] ( 0.040) ( 0.040) ( 0.038) You go to other relative in order to talk about job opportunities [yes=1] 0.124 0.080 *** 0.080 *** 0.080 *** 2348 [ 0.329] ( 0.028) ( 0.028) ( 0.026) You go to religious leader in order to talk about family problems [yes=1] 0.023 -0.003 -0.003 -0.003 2348 [ 0.150] ( 0.012) ( 0.012) ( 0.011) You go to religious leader in order to talk about conflicts with spouse/boyfriend [yes=1] 0.004 0.008 0.008 0.008 2348 [ 0.065] ( 0.005) ( 0.005) ( 0.005) You go to mentor in order to talk about work problems [yes=1] 0.007 0.007 0.007 0.007 2348 [ 0.082] ( 0.006) ( 0.006) ( 0.006) Member of professional association [yes=1] 0.038 0.028 0.027 0.026 2277 [ 0.192] ( 0.019) ( 0.019) ( 0.019) Member of group for girls [yes=1] 0.218 -0.076 * -0.075 * -0.077 ** 2258 [ 0.413] ( 0.039) ( 0.039) ( 0.039) Someone outside your family advises you on business [yes=1] 0.452 -0.010 -0.010 -0.005 2292 [ 0.498] ( 0.043) ( 0.043) ( 0.041) Parents support on wage employment index [1-5] 1 4.618 0.187 * 0.190 * 0.200 ** 2284 [ 0.929] ( 0.098) ( 0.098) ( 0.100) Parents support on self-employment index [1-5] 1 4.858 -0.011 -0.012 -0.007 2277 [ 0.580] ( 0.061) ( 0.061) ( 0.062) Controls No Yes No Using individual fixed effects No No Yes For all tables: Standard deviation is in brackets. Robust standard error is in parentheses. Level of Significance: * p<0.1; ** p<0.05; *** p<0.01. Notes: The variables used as controls in the regressions are engagement in income-generating activities, dummies for groups of age and education of respondents, household size and the number of female members in the household, school assistance, an indicator of whether the individual has children and marital status of the respondent. All controls are from baseline data. 1 The five categories of these variables are 1 for "Very disapproving," 2 for "Somewhat disapproving," 3 for "Neither supportive nor unsupportive," 4 for "Somewhat supportive," and 5 for "Very supportive."