Policy Research Working Paper 10924 Exploring the Drivers of Youth Pursuing Vocational Training in High-Paying Sectors in Côte d’Ivoire Clara Delavallade Manil Zenaki Léa Rouanet Estelle Koussoubé Africa Region A verified reproducibility package for this paper is Gender Innovation Lab available at http://reproducibility.worldbank.org, September 2024 click here for direct access. Policy Research Working Paper 10924 Abstract Education and skills are two key determinants of earning in these high-paying sectors. For women, previous train- potential, with sector specialization significantly influencing ing in similar fields strongly predicts their training choices, earnings. This study examines the drivers behind training highlighting path dependency. Additionally, women benefit choices in two high-paying sectors: information and com- more from male role models, which significantly increase munications technology (ICT) and energy. Drawing on their likelihood of choosing a training in ICT or energy. data from 2,528 individuals seeking vocational training in Women with greater agency are also more likely to opt for Côte d’Ivoire, we find that a majority (72% of men and training in these sectors. Conversely, women holding more 51% of women) aspire to train in ICT or energy. For both traditional views on specific household responsibilities are genders, higher levels of education and larger professional less likely to choose high-paying sector training. networks are positively correlated with selecting training This paper is a product of the Gender Innovation Lab, Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at mzenaki@worldbank.org. A verified reproducibility package for this paper is available at http://reproducibility.worldbank. org, click here for direct access. RESEA CY LI R CH PO TRANSPARENT ANALYSIS S W R R E O KI P NG PA The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Exploring the Drivers of Youth Pursuing Vocational ote d’Ivoire Training in High-Paying Sectors in Cˆ ∗ Clara Delavallade Manil Zenaki ea Rouanet L´ e Estelle Koussoub´ JEL Classification: I26, J16, L26, O15 Keywords: High-paying sectors, skills, vocational training, youth employment, gender and skills. ∗ Delavallade: World Bank, cdelavallade@worldbank.org; Zenaki: World Bank, mzenaki@worldbank.org; Rouanet: World Bank, lrouanet@worldbank.org; Koussoub´ e: World Bank, mkoussoube@worldbank.org. The order of author names was randomly assigned using the American Economic Association’s author randomization tool. This paper is a product of the World Bank Africa Gender Innovation Lab, Office of the Chief Economist, Africa Region. This project was conducted in collaboration with the International Rescue Committee and the PRO Jeunes (Pro-Youth) program. The authors would like to thank Patrice Comoe Boa, Michel Pokoudiby, Kevine Koffi Zoukou, Simplice Konan, Marie France Guimond, Natalia Strigin, Sreelakshmi Papineni and Girum Abebe Tefera for their invaluable contributions. We also extend our thanks to Innovations for Poverty Action (IPA) for their outstanding support in conducting the field experiment and data collection. Special thanks to Jos´ephine Tassy and Helene Donnat for their excellent research assistance. Additionally, we acknowledge Catherine Seya for her invaluable input and support throughout the research process. Lastly, we would like to express our sincere appreciation to the principal investiga- tors of the main experiment related to this study: Jeannie Annan, Dave Evans, and Markus Goldstein. Funding for this research was graciously provided by the Wellspring Foundation and the World Bank Umbrella Fund for Gen- der Equality (UFGE). The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Devel- opment/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. 1 Introduction Skills play an essential role in the economic transformation of nations, diverting labor away from low-productivity activities, such as subsistence farming and non-agricultural self-employment, to higher productivity occupations. This shift enables workers to increase their incomes within their ote d’Ivoire, as in many Sub-Saharan current jobs or transition to more lucrative occupations. In Cˆ African countries, a substantial proportion of students drop out during secondary school. While the gross enrollment rate in primary education approached 93% in the year 2022, this figure drops to 55% at the secondary education level. The high number of young people not engaged in formal education exacerbates the skills gap in the labor market. In this context, Technical and Vocational Education and Training (TVET) plays a pivotal role, not only in increasing the reservoir of relevant skills but also in providing a pathway to formal and better-paying employment opportunities.1 Both education and skill acquisition are strongly linked to higher income potential and improved labor market outcomes (Angrist and Krueger (1991), Psacharopoulos and Patrinos (2018), Kuepie et al. (2009), Alfonsi et al. (2020), Heckman et al. (2006)). Moreover, the specific sector of specialization can substantially influence earning potentials, with significant disparities in remuneration across various sectors (Goldstein et al. (2019), Campos et al. (2015)). While men often gravitate towards better-paying sectors such as Science, Technology, Engineer- ing and Mathematics (STEM), women tend to make different choices (Elu, 2018). Globally, less than one-third of women in higher education are in STEM-related fields, with this proportion even ote d’Ivoire, the context of this study, only 16% of women in lower in Sub-Saharan Africa. In Cˆ higher education study life sciences, mathematics or statistics (UNESCO, 2017). Similar imbalances can be observed in the TVET system. Despite achieving gender parity in TVET enrollment, includ- ote d’Ivoire. ing women in high-paying, male-dominated sectors (MDSs) remains a challenge in Cˆ Women’s enrolment is predominantly in services streams (89.9%), with limited representation in the secondary sector (10.2%), and almost no presence in agricultural streams (0.1%). These decisions have far-reaching consequences, as MDSs are often the most lucrative sectors.2 In many countries, women are concentrated in the least profitable sectors (World Bank (2022), Goldstein et al. (2019), Jonathan et al. (2015), Bardasi et al. (2011), Das and Kotikula (2019), (Alibhai et al., 2017)). Consequently, there is growing interest in supporting women to enter better- 1 The role of TVET in secondary education is still limited in Cˆote d’Ivoire. Despite an increase in enrolments in recent years, only 2 percent of the 15/24 years old participated in a TVET in 2018, according to UNESCO TVET Country Profile and UNESCO Institute for Statistics (UIS). 2 Our analysis of two nationally representative datasets, the EHCVM 2018 (WAEMU Commission, 2018) and the ENV 2015 (Institut National de la Statistique, 2015), reveals that MDSs offer higher average compensation compared to other sectors. Results are available upon request. 1 paying, often male-dominated, sectors. Acquiring the right skills through TVET can help women achieve this goal. UNESCO’s current TVET strategy3 explicitly emphasizes the need to develop targeted measures for inclusion and gender equality (UNESCO, 2022). Several studies have shown that women entrepreneurs are more likely to succeed if they receive technical training or participate in internships or apprenticeships(World Bank, 2022). Vocational training not only helps beneficiaries acquire new skills(Campos et al., 2018) but also build women’s self-confidence and challenge norms about gender-appropriate sectors (Croke et al. (2017), Alibhai et al. (2017)). These norms are often a significant barrier, preventing women from reaching their full potential (Das et al., 2023). Moreover, training provides opportunities to network in sectors that are typically less accessible to women and, under certain conditions, offers role models and mentors operating in better-paying sectors CLafortune et al. (2018),Sevilla et al. (2023)). By identifying with these role models and receiving their support, transitioning from one sector to another may seem more achievable. Vocational training can thus play a crucial role in reducing the gender pay gap. Despite the significant benefits, why do so few women seek training and education in sectors with higher earning potential? What distinguishes the women who do? Are their motivations different from those of men? These questions are at the heart of this paper. Understanding what drives women in Sub-Saharan Africa to pursue training in more lucrative sectors is crucial, yet the literature on this topic remains scarce. Gassier et al. (2022) show that increasing the salience of trade-specific returns nudges women with relevant technical knowledge and experience into seeking training in more lucrative trades. Buehren and Van Salisbury (2017) highlight the importance of networks, prospects, and personal preferences. Studies in other regions show that stereotypes and gender norms can create gaps between men and women in self-assessed skills and career aspirations. (Wang and Degol (2017), Reuben et al. (2014), Sansone (2017)). Addi- tionally, overconfidence and preferences for competitiveness and risk-taking drive gender differences in college major choices and earnings expectations (Reuben et al. (2017), Bench et al. (2015)) Despite the limited literature on this specific topic, we can draw insights from broader TVET lit- erature (Ayub (2017), Agodini et al. (2004)). This literature identifies several factors that influence training choices, including individual socio-demographic characteristics, education, past training, work experience, earnings, networks, role models, support systems, attitudes about gender roles and agency more broadly (UNESCO (2012), Arias et al. (2019), British Council (2020)). 3 ln line with national development priorities, UNESCO works to enhance the relevance of TVET to equip all youth and adults with the skills required for employment, decent work, entrepreneurship and lifelong learning. UN- ESCO’s efforts focus on three key areas outlined in its global Strategy for TVET (2022-2029): 1. Skills for indi- viduals to learn, work and live; 2. Skills for economies to transition towards sustainable development; 3. Skills for inclusive and resilient societies. 2 Building on these findings, this study explores the drivers of vocational training choices in two high-paying sectors: energy and information and communications technology (EICT). Both sectors were identified as male-dominated and high-paying sectors using a nationally representative dataset, the ENV 2015 (Institut National de la Statistique, 2015). Our empirical strategy uses linear regres- sions with cohort and city fixed effects to examine the correlates of training choices. The analysis is based on a sample of 1128 women and 1400 men who applied for job training in urban areas of ote d’Ivoire (Abidjan and Grand-Bassam). We use data collected as part of the PRO-Jeunes pro- Cˆ gram (or PRO-Youth in English). PRO-Jeunes was designed to provide (self-)employment support services that are particularly flexible to meet a wide range of beneficiaries’ needs. It targets youth aged 15 to 30 in rural and urban areas. During the application process, each candidate could choose between two main tracks: an employment track, which provided support for finding employment, or a self-employment track, which offered either a generalist path or a vocational training path. The generalist path was designed for candidates who were already working or engaged in an income- generating activity within a specific sector, or who had a particular sector in mind. For those opting for the vocational training path provided as part of the program, candidates could choose to enhance their skills in high-paying sectors such as energy or information and communications technology (ICT), or in retail, specifically in the sales of Lipton products. This study identifies several key factors influencing individuals’ decisions to pursue training in the EICT sectors. A significant portion of the sample, 72% of men and 51% of women, chose this path. Education and training are pivotal: an additional year of education increases the likelihood of seeking EICT training by 3.2 percentage points for women and 4.4 percentage points for men. This is partly due to the stringent educational entry requirements of vocational training programs. For women, prior training in EICT sectors markedly boosts the likelihood of pursuing further training in these high-paying fields, indicating path dependency. Additionally, a larger professional network increases the probability of choosing EICT training for both men and women by 3 percentage points. Gender-specific factors also play a role. Women benefit significantly from role models, particu- larly male role models, while men do not experience the same effect. Women with greater agency are more likely to opt for EICT training. However, traditional gender role attitudes can hinder women’s participation in EICT training. Women who prioritize household responsibilities are significantly less likely to choose EICT training. The main contribution of this paper is to demonstrate that different factors motivate men and women to train in high-paying, often male-dominated sectors. This reflects broader gender disparities in society. Our findings highlight the need for targeted interventions to address women’s 3 unique challenges in these fields. Effective approaches should focus on enhancing women’s skills, providing exposure to better-paying sectors, connecting them with supportive role models, and challenging prevailing gender norms and attitudes. 2 Context 2.1 ote d’Ivoire Labor Market In Cˆ Data from the World Bank4 sheds light on gender dynamics within the labor market of Cˆ ote d’Ivoire. As of 2023, the labor force participation rate for females stands at 56.5%, compared to 72.2% for males. These disparities also apply to young people aged 15 to 24. Notably, female labor force participation has shown an upward trend since 1990, indicative of a gradual narrowing of the gender gap. On a regional scale, the labor force participation rate for women in Sub-Saharan ote d’Ivoire’s 56.5%. The incidence of vul- Africa is reported at 60.7%, marginally higher than Cˆ nerable employment, characterized by informal work arrangements lacking social protection and safety nets, remains a significant issue. In 2022, 80.8% of employed women and 63.4% of employed ote d’Ivoire were engaged in vulnerable employment. Despite high rates, there has been men in Cˆ a discernible improvement in reducing vulnerable employment among females since 1991, reflecting ongoing structural changes within the labor market. Business ownership also reveals pronounced ote d’Ivoire was 26%. This gender disparities. In 2022, the proportion of female business owners in Cˆ underscores the continuing challenges women face in achieving parity in entrepreneurial activities. ote d’Ivoire face significant barriers to entrepreneurship, including limited access to Women in Cˆ finance and business networks. These barriers are exacerbated by socio-cultural norms that often discourage female entrepreneurship. Additionally, women’s participation in high-growth sectors re- mains low, further limiting their economic opportunities (The African Development Bank Group, ote d’Ivoire has made strides in increasing female labor force partic- 2021). In summary, while Cˆ ipation and reducing vulnerable employment among women, significant gender disparities persist, particularly in business ownership. Future policy interventions should aim to address these gaps to foster a more inclusive and equitable labor market. 2.2 ote d’Ivoire TVET In Cˆ According to the UNESCO Institute for Statistics (UIS) database, only 5% of secondary ed- ucation students are enrolled in vocational programs. At the same time, individuals with low 4 World Bank Gender Data Portal. 4 educational achievement have limited opportunities to access post-school training (Christiaensen and Premand, 2017). In 2018, just 2% of 15-24-year-olds participated in technical and vocational ote d’Ivoire. This low representation is partly due to insufficient capacity in public programs in Cˆ sector training institutions, which can only admit around 35,000 students annually (International Labour Organization, 2019). Moreover, the infrastructure and equipment in these institutions are in poor condition. To improve the quality of Technical and Vocational Education and Training (TVET), there is a need for revised and updated curricula, modernized equipment and facilities, teacher training with mandatory immersion courses, and better management of course information, anticipation of job market needs, and certification. Currently, TVET programs do not align with in- dustry requirements. Although connections with the private sector have been established, it remains challenging for graduates and school leavers to secure employment or internships. Frequent changes in governance and a lack of coherent long-term policies have further hampered the development of ote d’Ivoire. TVET in Cˆ The TVETs offered by PRO-Jeunes differ from most other programs in that they offer quality training focused on the technical skills needed for the job market. 2.3 PRO-Jeunes 2.3.1 Enrollment in the PRO-Jeunes program ote d’Ivoire (Abidjan, the country’s capital and We rely on baseline surveys collected in urban Cˆ Grand Bassam, the historic capital situated at around 33 km from Abidjan). These surveys were conducted as part of the impact evaluation of the PRO-Jeunes (Pro-Youth ) program, a six-year program (2017-2022) implemented by the International Rescue Committee (IRC). The program provides flexible (self-) employment support services to youth aged 15 to 30 years, in both rural and ote d’Ivoire.5 In addition, to be eligible for the PRO-Jeunes program, applicants must be urban Cˆ unschooled or have dropped out of school, and show proof of identity. The program was designed to accommodate a wide range of beneficiaries’ needs and knowledge levels, with one cohort targeted approximately every six months for a total of eight cohorts. For this study, we focus on two of the eight cohorts. The Pro-Jeunes initiative was promoted through a comprehensive communication strategy coor- dinated by the IRC team, aimed at optimizing beneficiary recruitment and enhancing the program’s visibility. This strategy employed a blend of public relations, direct marketing, media outreach, and 5 The age range of youth targeted by the project was expanded from 15-24 years old for the first three cohorts to 15-30 years old to take into account context specific constraints to youth recruitment into the project. 5 community mobilization efforts. Key activities included institutional campaigns targeting local au- thorities, vocational training centers, employment agencies, social centers, and private enterprises. Media outreach involved press releases, advertisements, and direct broadcasts on the Pro-Jeunes website, as well as social media platforms (X, YouTube, Facebook) and local and regional television and radio outlets. These actions were strategically designed to inform the general public and engage specific target audiences. The core messaging of the campaign positioned Pro-Jeunes as an innovative solution providing e- learning, blended learning, and technical training, with the goal of professionally integrating 10,000 youths, with a strong emphasis on young women. Community outreach and mobilization included meetings with community leaders, distribution of informational flyers, and communication with communities about the project launch and pre-enrollment sites. Two pre-enrollment methods were implemented. The first involved applicants visiting fixed Pro- Jeunes pre-enrollment centers, located in partner organization premises within neighborhoods and Pro-Jeunes offices. The second method involved outreach by agents who visited targeted populations directly. Caravans were deployed in neighborhoods and villages with high concentrations of potential beneficiaries, identified based on a pre-established map. These caravans operated during specific communication and pre-enrollment campaign periods (January to March, and July to September each year), with mobile enrollment points established in areas distant from the fixed sites. Interested youths completed registration forms to express their intent to participate in the program and were required to provide additional documents (ID and a letter of motivation) to finalize their application. The IRC team collected and screened this information to assess the eligibility of applicants. Eligible candidates were then contacted to provide further details about the program and confirm their continued interest. Finally, those who confirmed their interest were invited to attend an enrollment meeting to formalize their participation. 2.3.2 Training paths Eligible candidates who confirmed their interest must choose a training path. Figure 1 shows the different training tracks and the number of men and women in each track. Eligible candidates can opt for either the wage-employment pathway or the entrepreneurial one. For those who select entrepreneurial training (95% of our sample), two options are available. They can choose the gener- alist path or receive vocational training in one of the three following sectors: energy, trade/retail, or information and communications technology (ICT). It is worth noting that energy and ICT (EICT) 6 are traditionally MDSs.6 In total, 72% of men and 51% of women chose EICT. Individuals who opt for the wage-employment path are coached to find a salaried job. After a skill assessment, they formulate a professional project and receive help writing CVs and cover letters. In addition, they are trained for job interviews and receive help in researching job opportunities. Candidates who choose the generalist entrepreneurship track are encouraged to develop a business plan and receive continued coaching for 3 to 6 months. They are also linked to financial services and participate in business competitions. Individuals who opt for vocational training benefit from a different curriculum depending on their chosen path. PRO-Jeunes Energy aims to train technicians either in domestic electrical installation or in the design and installation of solar photovoltaic systems for applications such as buildings, solar pumping, solar street lamps, etc. Projeunes Energy is a 6-month training that requires a minimum education level of 7 years and to pass a small exam given by IRC. Courses take place in training centers equipped with electrical equipment and materials. This training program has Schneider Electric as its technical partner. PRO-Jeunes TIC offers a 2-month training course in digital marketing. It requires at least 8 years of education. Finally, the retail track (PRO-Jeunes Lipton) provides training in sales and negotiation techniques. This course requires a minimum level of education of 3 years and has Unilever as its technical partner. It is important to note that vocational training can be seen as a more attractive pathway than others, as it involves recognized technical partners and can therefore more readily lead to job opportunities. Individuals who have completed vocational training are not precluded from seeking paid employment following their training. 6 In addition to the age requirement, candidates interested in applying for vocational training should have a min- imum educational level of 3 years for the trade/retail path, 7 years for the Energy path, and 8 years for the ICT path 7 Figure 1: PRO-Jeunes training tracks Notes: This figure shows the different training tracks of the PRO-Jeunes program. In addition, the figure indicates the proportion of men and women in our sample in each track. 3 Data, Measures, and Sample Description 3.1 Data collection Our study draws on two baseline surveys (corresponding to two enrollment cohorts) conducted ote with potential beneficiaries of the PRO-Jeunes program in Abidjan and Grand-Bassam, Cˆ d’Ivoire. The first survey, conducted between February and July 2019, included 1,865 respondents, while the second survey conducted between May and September 2020, included 1,210 individuals. The first survey was carried out in person, while the second was carried out by telephone due to the COVID-19 pandemic. In these surveys, we collected information on household characteristics and assets, as well as extensive respondent-level information on education, training, agency, aspirations, and employment in the past 30 days. We also collected data on respondents’ networks and role models, as well as their attitudes toward gender roles and domestic violence. 8 3.2 Measures 3.2.1 High-paying sectors and sectors dominated by men We asked each respondent whether they engaged in an income-generating activity within the 30 days preceding the survey. For those who answered affirmatively, we collected information on the nature of their employment (self-employed or salaried), the type of activity or sector, and the in- come generated. We classify a sector as male-dominated when at least 75% of its workforce is male. ote d’Ivoire, we drew data from the ENV To establish the proportion of men in each sector in Cˆ 2015 (Institut National de la Statistique, 2015). The ENV is a household living standards survey ote d’Ivoire. The survey interviewed conducted in 2015 by the National Institute of Statistics in Cˆ 47,635 individuals from 12,900 households and is also nationally representative. It is worth noting that six occupations could not be categorized, and we presumed they were not male-dominated by default. 11% of respondents engaged in unclassified income-generating activities.7 The resulting classification is summarized in Table 1 for the working sectors. Mining/oil worker, electrical, refrig- eration and air conditioning belong to the energy sector while computer and electronics belong to ICT. To ensure that our classification of MDSs is not highly dependent on the threshold used, we verified the robustness of our classification in Table A.1 and in Figure 9, in Appendix A. In addition, Figure 2 provides an overview of the energy and ICT sectors compared to others in terms of earnings and proportion of men. 7 We also use the EHCVM 2018 dataset (WAEMU Commission, 2018), and find that three out of the six un- classified sectors are indeed not male-dominated, resulting in less than 1% of respondents engaging in unclassified activities. Results are available upon request. 9 Table 1: Male dominated working sectors Activity/Sector Category Male percentage Median earning PPP Chemical technician MDS 100 1197 Piping/Boil making MDS 100 718 Real estate activity MDS 100 299 Cobbling MDS 100 279 Welding MDS 100 259 Metal joining MDS 100 239 Fishing MDS 100 199 Security guard MDS 100 199 Refrigeration and air conditioning [Energy] MDS 100 180 General mechanics MDS 99.83 199 Transportation of people (bus/taxi) or goods MDS 99 299 Auto mechanics MDS 97.7 239 Painting MDS 97.04 299 Material handling MDS 96.87 479 Furniture making MDS 96.42 279 Plumbing MDS 95.77 239 Priest/Pastor/Imam/Marabout MDS 92.44 199 Sports MDS 92.31 199 Computer/Electronics/computer and mobile repair [ICT] MDS 87.78 339 Electricity [Energy] MDS 85.72 359 Artist/art Maker/photographer/musician MDS 83.49 399 Teaching and education professions MDS 82.12 399 Construction (masonry/scaffolding) MDS 81.47 180 Mining/oil worker [Energy] MDS 80.42 319 Pump operator (gas station) MDS 76.56 319 Liberal intellectual and scientific professions(journalists, MDS 75.55 479 lawyers, accountants, etc.) Tutor (private teacher) Not MDS 74.14 798 Factory worker/industrial manufacturing Not MDS 73.71 319 Sewage and refuse disposal Not MDS 73.16 160 Public administration Not MDS 68.58 598 Political party leader/executive Not MDS 67.16 120 Agriculture/livestock Not MDS 66.84 100 Laundry/Textile Cleaning Not MDS 65.87 100 Textile: sewing/embroidery/weaving/dyeing Not MDS 53.12 140 Health and social work professions (doctors, pharmacists, Not MDS 48.58 287 nurses, other) Retail trade: other Not MDS 41.35 160 Services in the hotel/restaurant industry Not MDS 27.22 180 10 Table 1 continued from previous page Activity/Sector Category Male percentage Median earning PPP Maintenance and cleaning/surface technician Not MDS 22.92 80 Retail trade: foodstuffs Not MDS 21.39 140 Hairdressing/cosmetic care Not MDS 20.78 140 Governor Not MDS 17.93 180 Food/beverage processing Not MDS 9.77 120 Kitchen/pastry/restaurant Not MDS 5.54 239 Other* Not MDS NA NA Miscellaneous services (campaigns, events, etc.)* Not MDS NA NA Chair/chapel rental* Not MDS NA NA Video game room* Not MDS NA NA Discotheque* Not MDS NA NA Telephone booth* Not MDS NA NA N otes: *Not MDS by default. [ICT] and [Energy] shows sectors that were assigned respectively to the ICT and Energy sectors. The sector classification is based on data from the ENV 2015 survey, which includes information on respondents’ main activities and their corresponding sectors. Given that the Pro-Jeune survey uses its own sector classification system, differing from the ENV classification, we matched the ENV sectors to those of the Pro-Jeune survey. 11 Figure 2: Scatterplot of median earnings on the proportion of men working in each trade using the ENV 2015 dataset 3.2.2 Characteristics potentially associated with training choices We aim to identify the factors associated with training choices in high-paying sectors and have organized our variables into six categories based on the existing literature: (i) socio-demographic characteristics, (ii) education and training, (iii) employment and earnings, (iv) network, (v) role model and support, (vi) agency and gender attitudes.8 All variables are listed in Table 2. Specifically, we consider the respondent’s age, number of dependent children, and wealth decile (top 40% of households with the most assets) as socio-demographic variables. In terms of education and training, we consider whether respondents have attended any qualifying training, the field of training and whether the training was in EICT. We also analyzed the relationship between training choices and variables related to employment and earnings. These variables include information on the type of employment (salaried or self- employed), the type of activity/sector, and the income generated. Additionally, we examined the size and composition of each respondent’s professional network by asking them to name 3-5 people they could count on in their professional life. We recorded the relationship to the contact (friend, family, other) and their professional activity. We also asked respondents if they knew someone 8 We also explored the correlation between social-emotional skills and training choices, but our results did not show a significant relationship. The effect was very close to zero and smaller than the minimum detectable effect. These results contrast with (Das et al., 2023). Results are available upon request. 12 who had been successful in their professional life, i.e., a role model, and noted the relationship with the role model and the role model’s occupation for those who answered in the affirmative. In addition, we asked respondents if they had access to individuals outside of their family for work- or business-related advice, and if so, the gender of the designated individual(s). The survey included modules measuring respondents’ agency, attitudes toward gender roles, and attitudes toward domestic violence (Table A.2 contains the full list of questions, with potential answers). To assess attitudes toward gender roles, we read four sentences that reflected common stereotypes about men’s and women’s abilities and roles. A score aggregating answers to all the questions was then computed. Finally, we calculated a cohort-specific z-score. To construct our agency measure, respondents were asked about their ability to make decisions in six different areas: entrepreneurial activity, employment, purchasing durable goods, minor household expenses, health, and daily tasks. A z-score by cohort was again calculated to reflect the level of agency. Finally, to assess attitudes toward domestic violence, participants were asked whether they believed it was justified for men to hit or beat their wives in four scenarios. The sum of affirmative responses was converted into a cohort-specific z-score to obtain the final score. 13 Table 2: Variables included in the analysis by category Variables Isolated factors Combined factors 1- Sociodemographic characteristics Age of the respondent (z-score) 1 1 Number of dependent children (z-score) 1 1 Household wealth index by cohort (z-score) 1 1 2- Education and training Years of education (z-score) 1 1 Any training 1 0 Already received training in EICT in the past 1 1 3- Employment and revenues Wage-employed in the last 30d 1 0 Self-employed in the last 30d 1 0 Had a paid work in the last 30d 1 1 Worked in EICT in the last 30d 1 1 Revenues earned in the last 30d (z-score) 1 0 4- Network Has a professional Network 1 0 Network size (z-score) 1 1 Proportion of males in the network (z-score) 1 1 5- Role model and support Male role model 1 1 Female role model 1 1 Role model in EICT 1 0 Can ask for professional advice from people around him/her 1 1 (outside the family) 6- Gender attitudes and agency Gender attitudes (z-score) 1 1 Agrees that women’s most important role is to cook and take 1 0 care of her household Agrees that household expenses are the responsibility of the hus- 1 0 band Agrees that by nature men and women have different abilities in 1 0 differenta areas Agrees that at work, men cope better with difficult conditions 1 0 than women 14 Table 2 continued from previous page Variables Isolated factors Combined factors Agency: input in productive decisions (z-score) 1 1 Attitude towards domestic violence (z-score) 1 0 N otes: Variables included in the analysis by category. Categories are shown in bold. A value of 1 in the second col- umn signifies that the variable was selected for the isolated factors analysis, while a value of 1 in the third column indicates its selection for the combined factors analysis. 15 3.3 Sample description Our sample consists of individuals for whom we have information on all six previously estab- lished categories of variables: (i) socio-demographic characteristics, (ii) education and training, (iii) employment and income, (iv) network, (v) role model and support, (vi) agency and gender attitudes. We interviewed 3075 people, but our analysis sample consists of only 2528. Any respondent with at least one missing value for any of the above groups of variables is excluded from the analysis. Excluded respondents differ slightly from their counterparts: on average, they are one and a half years older and more literate. In addition, they are more likely to be a woman from Abidjan, to have received training in EICT in the past, and to have a male role model. Of the 2528 respondents included in our sample, 5% opted for wage-earning training, 30% for general entrepreneurship training, and 65% for vocational training. Of those choosing vocational training, 97% opted for EICT. Table 3 presents descriptive statistics by gender for individuals from the cities of Abidjan and Grand Bassam separately and for the entire sample. We also examine differences between men and women in each city. Our sample is well-balanced in terms of gender, with nearly equal numbers of men and women. The characteristics of male and female respondents are very similar in Abidjan and Grand Bassam. However, the much smaller sample size in Grand Bassam induced a limited power to detect gender differences. Most respondents are from Abidjan. Women are, on average, 22 years old, the same as men, but have more children. Women have completed one year less education than men, earn less income, and come from households with fewer assets (significant at least at 5%). Additionally, a higher proportion of women have not had paid work in the last 30 days. As might be expected, men are more likely to have received training in EICT (a male-dominated sector) and to be working there. Of the men, 19% worked in EICT in the month before the survey, compared to 6% for women. In addition to having a smaller professional network, women also have a smaller proportion of male contacts in their network. This may imply fewer job opportunities in sectors dominated by men. In line with previous literature, men are significantly more likely to agree with most conservative statements on gender norms, such as ”men and women have different abilities in different areas” or ”women’s most important role is to cook and take care of her home”. In contrast, women show a higher tendency to justify domestic violence. This may stem from the fact that they are less educated and come from poorer households, as suggested in other studies (Doku and Asante (2015), Yount and Li (2009)). It is worth noting that despite the barriers they face, more than half of the women in our sample 16 expressed a desire to pursue professional training in EICT, which are known to be higher-paying sectors. Thus, the absence of women from these sectors is not always a matter of choice. 17 Table 3: Descriptive Statistics Abidjan Grand Bassam Sample Male Female Diff Male Female Diff Male Female Diff PRO-Jeunes Training choice Wage-employement track 0.05 0.05 -0.01 0.04 0.04 0.00 0.04 0.05 -0.01 Generalist entrepreneurial track 0.20 0.37 -0.17** 0.37 0.60 -0.23 0.23 0.41 -0.18*** Vocational training track 0.76 0.58 0.18** 0.59 0.36 0.23 0.73 0.54 0.19*** Vocational training in energy 0.50 0.33 0.17*** 0.56 0.34 0.23 0.51 0.33 0.18*** Vocational training in information and communication 0.25 0.22 0.03 0.01 0.00 0.01 0.21 0.18 0.03 technologies (ICT) Vocational training in trade 0.01 0.03 -0.02 0.01 0.03 -0.01 0.01 0.03 -0.02* Sociodemographics Age of the respondent 22.41 22.46 -0.05 21.42 20.85 0.57 22.25 22.19 0.06 SD 3.07 3.35 2.81 2.94 3.06 3.34 18 Number of dependent children 0.34 0.67 -0.33*** 0.19 0.79 -0.61** 0.31 0.69 -0.38*** SD 0.80 0.96 0.58 1.09 0.77 0.98 Household is in the top 2 wealth quintile (by cohort) 0.44 0.38 0.06** 0.33 0.32 0.02 0.42 0.37 0.05** Education and training Years of education 10.74 9.74 1** 9.12 8.28 0.84 10.48 9.50 0.98*** SD 3.20 3.99 3.68 3.93 3.34 4.02 Any training 0.23 0.23 0.00 0.23 0.28 -0.06 0.23 0.24 -0.01 Already received training in EICT in the past 0.11 0.09 0.02 0.09 0.05 0.04 0.11 0.09 0.02 Want to be trained in energy or ICT 0.75 0.55 0.20** 0.58 0.34 0.24 0.72 0.51 0.21*** Employment and earnings Wage-employed in the last 30d 0.68 0.43 0.26** 0.60 0.51 0.09 0.67 0.44 0.23*** Self-employed in the last 30d 0.35 0.40 -0.04 0.29 0.47 -0.18 0.34 0.41 -0.07* Had a paid work in the last 30d 0.72 0.53 0.19** 0.65 0.60 0.05 0.71 0.54 0.17*** Worked in EICT in the last 30d 0.20 0.05 0.15*** 0.13 0.07 0.06 0.19 0.06 0.14*** Revenues earned in the last 30d (USD PPP) 290.02 109.27 181*** 221.54 118.57 103 279.01 110.84 168*** SD 391.64 184.95 320.57 241.02 381.84 195.43 Network size and characteristics Table 3 continued from previous page Abidjan Grand Bassam Sample Male Female Diff Male Female Diff Male Female Diff Has a professional Network 0.82 0.75 0.07* 0.71 0.82 -0.11 0.80 0.76 0.04 Network size (top coded at 3) 1.91 1.65 0.26** 1.83 1.81 0.02 1.90 1.68 0.22** SD 1.35 1.32 1.49 1.23 1.38 1.31 Proportion of male contacts in the top 5 contacts 0.55 0.42 0.13** 0.54 0.43 0.11 0.55 0.42 0.13*** SD 0.41 0.41 0.42 0.38 0.41 0.40 Role Model and support outside the family Male role model 0.72 0.34 0.38*** 0.66 0.33 0.33 0.71 0.34 0.37*** Female role model 0.07 0.34 -0.27*** 0.09 0.28 -0.19 0.07 0.33 -0.26*** Role model in EICT 0.09 0.07 0.02 0.08 0.05 0.03 0.09 0.07 0.02 Can ask for professional advice from people around 0.84 0.74 0.10*** 0.80 0.81 -0.01 0.83 0.75 0.08*** him/her (outside the family) Gender attitudes and agency 19 Gender attitudes [0,1] 0.48 0.46 0.01 0.46 0.44 0.01 0.47 0.46 0.01 SD 0.18 0.18 0.16 0.16 0.18 0.18 Agrees that women’s most important role is to cook and 0.52 0.47 0.05 0.63 0.56 0.06 0.54 0.48 0.05* take care of her househol Agrees that household expenses are the responsibility of 0.70 0.61 0.09* 0.72 0.60 0.12 0.70 0.61 0.10** the husband Agrees that by nature men and women have different 0.85 0.78 0.07** 0.78 0.78 0.00 0.84 0.78 0.06** abilities in differenta areas Agrees that at work, men cope better with difficult condi- 0.34 0.37 -0.03 0.22 0.32 -0.10 0.32 0.36 -0.04 tions than women Perceived ability to take decisions alone if desired 0.72 0.67 0.05** 0.72 0.65 0.07 0.72 0.66 0.06** (score=[0,1]) SD 0.25 0.26 0.26 0.22 0.25 0.26 Attitudes towards domestic violence [0,1] 0.07 0.09 -0.02*** 0.07 0.11 -0.04 0.07 0.09 -0.02** SD 0.15 0.19 0.16 0.18 0.15 0.19 N 1175 938 225 190 1400 1128 Table 3 continued from previous page Abidjan Grand Bassam Sample Male Female Diff Male Female Diff Male Female Diff Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 This table shows the descriptive statistics, i.e., the characteristics of the sample. For each characteristic, we indicate the mean. When the variable is continuous, the standard deviation (SD) is shown in italics below the mean. The acronym EICT stands for Energy or Information Communication and Technology. Gender attitudes are measured as the ratio of stereotypes to which the respondent agrees, divided by the number of questions on gender attitudes. Hence, when the variable gender attitudes is equal to 1, the respondent agrees with all stereotypes. Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrap to compute p-values. 20 4 Empirical Strategy To investigate the factors influencing the decision to pursue vocational training in EICT, we employ a two-step approach: an individual factor analysis and a combined factor analysis. To complement this analysis and test the robustness of our main results, we use a LASSO procedure for variable selection. We first examine the correlation between the choice of training in EICT and each factor inde- pendently using the following specification: Yijc = β0 + β1 Xijc + γc + λj + ϵijc (1) Where Yijc is a dummy variable for whether respondent i, from city j , cohort c, chose vocational training in EICT. Xijc is a given factor of interest potentially associated with training choice as described in Section 3.2.2, while γc and λj are cohort and city fixed effects, respectively. We perform a separate regression for each factor X . The parameter of interest, β1 , indicates the association between a specific factor and vocational training choice in EICT. Results are shown in section 5.1. We then rely on the following specification to analyze the effects of all factors simultaneously: N ′ Yijc = β0 + βn Fnijc + γc + λj + ϵijc (2) n=1 ′ contains a subset of variables in variable category n. In each category, a subset The vector Fn of variables is selected based on literature and hypotheses about which are most likely to impact training choices. We focus on six categories, as detailed in Section 3.2.2, Table 2. These categories are introduced sequentially, starting with sociodemographic characteristics and ending with gender attitudes and agency. When all six variable categories are introduced simultaneously, N = 6. The vector βn includes the coefficients that indicate the association between each selected factor within the category n and the choice of vocational training in EICT. 5.2. To complement our analysis and address potential multicollinearity issues, we use a LASSO procedure to identify the most relevant predictors of vocational training choice in EICT. This provide a robustness check for our main results. Table B.3 shows the variables selected using the LASSO procedure alongside variables selected in the combined factor analysis. We estimate all equations separately for men and women to capture potential gender differences in factors influencing training choices. Standard errors are clustered at the zone level.9 9 Our sample includes six zones in total. Given the low number of clusters, we rely on wild cluster bootstrapping ecoub´ (Cameron et al., 2008). A zone corresponds to a group of municipalities. Municipalities included are: Att´ e, 21 5 Results In this section, we present the results of our analysis of factors influencing the demand for training in EICT. As mentioned previously, we examined six key categories of factors: sociodemographic characteristics, education and training, employment and income, networks, role models and support, and agency and attitudes toward gender roles and domestic violence. Our results are organized into two main subsections. We begin by presenting the findings from the analysis of each factor independently. Figures 3 through 8 show the coefficient estimates by factor category, separately for men and women.10 We then discuss the results of our analysis considering multiple factors simultaneously, high- lighting which factors remain significant predictors of training choices when examined together. 5.1 Isolated Factors In this initial step, we consider each factor individually to provide a clear picture of its potential influence on training choices. This step is important as it allows us to identify individual correlations and potential mechanisms before examining the effects of multiple factors simultaneously. • Socio-demographic characteristics: Women from relatively wealthier households are more likely to apply for training in EICT fields (Figure 3). A one standard deviation increase in the wealth index is associated with a 4 percentage point increase in the probability of opting for the EICT training. Conversely, the more children a woman has, the less likely she is to seek training in those high-paying sectors. Having more children decreases the probability of opting for EICT training by 4 percentage points. Note that this is not a driving factor for men. This is consistent with the fact that the burden of household chores and childcare falls most heavily on women (Cassirer and Addati (2007), Clark et al. (2019)). • Education and training: Minimum education levels and specific skills are prerequisites for certain sectors, particularly STEM-related professions. The PRO-Jeunes program illustrates this: applicants for vocational training need a minimum education level of 3 years for the trade/retail path, 7 years for the Energy path, and 8 years for the ICT path. Higher education levels before training correlate with increased success in the field. Accordingly, with more years Yopougon, Songon, Abobo, Anyama, Brofodoume, Adjame, Cocody, Bingerville, Plateau, Treichville, Marcory, Koumassi, Port-Bouet and Grand Bassam 10 Table C.4 in the Appendix provides these results in a table format, including tests for gender differences in coefficients. 22 of education our respondents are significantly more likely to choose EICT training (Figure 4). A one standard deviation increase in education increases the likelihood of seeking training in EICT by 14 percentage points for men and 15 percentage points for women. Moreover, those who have previously trained in EICT are significantly more likely to repeat the experience, with this effect being nearly three times stronger for women. • Employement and income: Self-employed women are 10.7 percentage points more likely to seek EICT training, while self- employed men are 5.2 percentage points less likely (Figure 5). For women, wage employment negatively correlates with choosing EICT training. • Network: We find suggestive evidence of the importance of network size for both genders (Figure 6). For women, a one standard deviation increase in network size corresponds to a 4.5 percentage point increase in the likelihood of seeking EICT training. For men, the effect is about half as strong but still significant. However, there is no significant correlation between the gender composition of the network and training choices for either gender. • Role models and support: Role models may serve as sources of inspiration and mentoring and can encourage individuals to enter specific fields of activity. Women with mentors, es- pecially male mentors (Figure 7), are more likely to seek vocational training in EICT (+19.1 percentage points). In addition, women who have access to professional advice from men and women outside their families are 10.6 percentage points more likely to pursue training in these high-paying sectors. Interestingly, men also benefit from female role models, especially those working in EICT, which increases their likelihood of training in the same sector (+10.3 percentage points). • Agency and attitudes toward gender roles and domestic violence: Interestingly, we find no significant correlation between agency (measured by decision-making power in personal and household matters) or attitudes toward domestic violence and EICT training choices. However, gender role attitudes show some significant associations (Figure 8). For instance, we observe a marginal correlation between our index of attitudes toward gender roles and training choices for men (significant at 10%). In addition, we find that specific attitudes can influence training choices for both men and women. In our sample, 48% of women and 54% of men believe that a woman’s primary role is to take care of the household and cook. Women who hold this belief are 18 percentage points less likely to choose training in EICT. Surprisingly, 23 men who hold this belief are also less likely to choose to train in EICT, though the effect is half as strong. Similar effects are observed for those who believe household expenses are solely the husband’s responsibility, with women being 11.2 percentage points less likely and men 4.2 percentage points less likely to opt for EICT training. While these findings suggest an influence of norms and beliefs on individuals’ training choices, it is important to consider other factors, such as education and skills, that may contribute to the observed associations. These results suggest that women face multiple barriers to entering high-paying, often male- dominated sectors. To further explore the interplay between these factors in influencing training choices in EICT, we now turn to our combined factor analysis. Tables 4 and 5 respectively. The factors considered are listed in Table 2. 5.2 Combined factors In this section, we present results showing the effect of all selected factors on training choices. Variables were selected based on literature and hypotheses about which are most likely to impact training choices. We introduce variables sequentially by factor category. When all factors are considered together (Tables 4 and 5), we observe changes in the correlations between factors and training choices. This underscores the complexity and interplay between factors influencing training choices in EICT. Notably, the previously observed negative correlation between the number of dependent children and women’s training choices becomes non-significant when we control for education (Table 4, column (2)). This suggests that other factors, like education, mitigate the influence of the number of dependent children on training decisions. The same goes for household wealth and gender attitudes (columns (5) and (6), respectively). It is important to note that despite the lack of significance of the overall gender attitudes index, we cannot exclude that specific beliefs about gender roles may be relevant. Conversely, the correlation between network size remains significant at 5% for men and 10% for women (+3 percentage points for both genders). This highlights the importance of how networks influence training choices. In addition, the influence of role models, especially male role models, on women’s training choices remains high (+16 percentage points) and significant at 10%. On the other hand, the ability to seek professional advice outside the family is no longer significant. This result should be interpreted with caution, as role models can be mentors who themselves provide advice. Similarly, women with prior training in EICT are 13 percentage points more likely to opt for the same sector. These findings emphasize the persistent significance of educational background and 24 prior training experiences in shaping women’s decisions to seek training in EICT. Finally, women with greater agency are more likely to opt for EICT. Overall, our analysis reveals a complex association between socio-demographic characteristics, education and training, employment, networks, role models and support, gender attitudes, and agency, and training choices, with some notable gender differences as well as similarities. For women, education, prior training in EICT, role models, and agency are predominant in shaping their training decisions. For men, education, age, and network characteristics (size and composition) significantly influence their training choices. Our LASSO analysis (Table B.3) used as a robustness check, further highlights these gender differences while also providing a more parsimonious set of predictors.11 For men, the LASSO selects only education as a significant predictor, suggesting it may be the most robust factor. For women, the LASSO identifies a broader range of factors, including age, household wealth, education, network size, male role models, and access to professional advice outside the family. 11 While the LASSO provides a more conservative estimate of significant predictors, the broader set of factors identified in our combined analysis offers valuable insights into the complex interplay between socio-demographic characteristics, education and training, employment, networks, role models and support, gender attitudes, and agency, and training choices, particularly for women 25 Figure 3: Correlation between the choice of training and sociodemographic characteristics Notes: Coefficient estimates from an OLS regression of training choice on socio-demographic char- acteristics as specified in equation 1, separately for women and men. For each variable, a separate regression is performed. Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrapping to compute confidence intervals. We use a standardized z-score for age, number of children and household wealth index. 26 Figure 4: Correlation between the choice of training, education, and past training Notes: Coefficient estimates from an OLS regression of training choice on education, and past train- ing as specified in equation 1, separately for women and men. For each variable, a separate regression is performed. Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrapping to compute confidence intervals. We use a standardized z-score for education. The other variables are dummies indicating whether the respondent received any training in the past or trained in EICT. The acronym EICT stands for Energy or Information and Communications Technol- ogy. 27 Figure 5: Correlation between the choice of training, employment and revenues Notes: Coefficient estimates from an OLS regression of training choice on employment characteristics and revenues, as specified in equation 1, separately for women and men. For each variable, a separate regression is performed. Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrapping to compute confidence intervals. All variables are dummies. The vari- able ”Wage-employed in the last 30d” indicates whether the respondent worked for someone during the last 30 days in exchange for a salary, a commission, or compensation. The ”Paid work in the last 30 days” variable indicates whether respondents have worked in the last 30 days and earned money as a result. The acronym MDSs stands for male-dominated sectors, while EICT stands for Energy or Information and Communications Technology. 28 Figure 6: Correlation between the choice of training and network Notes: Coefficient estimates from an OLS regression of training choice on network characteristics, as specified in equation 1, separately for women and men. For each variable, a separate regression is performed. Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrapping to compute confidence intervals.The variable ”Has a professional network” is a dummy. We use a standardized z-score for network size and proportions of males in the network. 29 Figure 7: Correlation between the choice of training, role models and support Notes: Coefficient estimates from an OLS regression of training choice on role models and support, as specified in equation 1, separately for women and men. For each variable, a separate regression is performed. Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrapping to compute confidence intervals. All variables are dummies. The acronym EICT stands for Energy or Information and Communications Technology. 30 Figure 8: Correlation between the choice of training, agency, gender attitudes and domestic vio- lence Notes: Coefficient estimates from an OLS regression of training choice on agency, attitudes toward gender roles and domestic violence, as specified in equation 1, separately for women and men. For each variable, a separate regression is performed. Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrapping to compute confidence intervals. We use a stan- dardized z-score for agency, gender attitudes and attitudes toward domestic violence (other variables are dummies). A higher score means a higher agency, conservative gender attitudes and a higher ten- dency to justify domestic violence. 31 Table 4: Correlates of women’s training choices in ICT and energy Chose the ICT or the energy vocational training VARIABLES (1) (2) (3) (4) (5) (6) Sociodemographic characteristics Age of the respondent (z-score) 0.018 -0.011 -0.012 -0.013 -0.014 -0.019 (0.013) (0.012) (0.012) (0.012) (0.014) (0.015) Number of dependent children (z-score) -0.046** -0.021 -0.022 -0.021 -0.022 -0.022 (0.010) (0.015) (0.015) (0.015) (0.015) (0.015) Household wealth index by cohort (z-score) 0.036* 0.032* 0.033* 0.029* 0.026 0.028 (0.017) (0.015) (0.015) (0.013) (0.013) (0.014) Education and training Years of education (z-score) 0.142* 0.142* 0.141* 0.130* 0.129** (0.015) (0.014) (0.015) (0.017) (0.018) 32 Already received training in EICT in the past 0.139** 0.141** 0.148** 0.133** 0.129** (0.032) (0.031) (0.036) (0.038) (0.036) Employment and revenues Had a paid work in the last 30d 0.029 0.026 0.026 0.023 (0.016) (0.016) (0.018) (0.019) Worked in EICT in the last 30d -0.010 -0.010 -0.014 -0.011 (0.055) (0.055) (0.051) (0.054) Network Network size (z-score) 0.034* 0.031* 0.034* (0.013) (0.013) (0.012) Proportion of males in the network (z-score) 0.005 0.002 0.000 (0.013) (0.012) (0.011) Role model and support Male role model 0.158* 0.159* Table 4 continued from previous page Chose the ICT or the energy vocational training (1) (2) (3) (4) (5) (6) (0.025) (0.028) Female role model 0.054** 0.052** (0.021) (0.018) Can seek professional advice from individuals outside 0.002 0.003 the family (0.041) (0.043) Gender attitudes and agency Gender attitudes (z-score) -0.030 (0.027) Agency: input in productive decisions (z-score) 0.021** 33 (0.011) Observations 1128 1128 1128 1128 1128 1128 Cohort FE YES YES YES YES YES YES City FE YES YES YES YES YES YES N otes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 . Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrap to compute p-values. For each column, we include cohort and city fixed-effects. The acronym EICT stands for Energy or Information Communication and Technol- ogy. Table 5: Correlates of men’s training choices in ICT and energy Chose the ICT or the energy vocational training VARIABLES (1) (2) (3) (4) (5) (6) Sociodemographic characteristics Age of the respondent (z-score) -0.011 -0.038** -0.037** -0.037** -0.038** -0.036** (0.013) (0.009) (0.010) (0.011) (0.011) (0.011) Number of dependent children (z-score) -0.006 -0.000 0.000 -0.002 -0.002 -0.002 (0.012) (0.012) (0.013) (0.012) (0.013) (0.013) Household wealth index by cohort (z-score) 0.031 0.015 0.013 0.010 0.010 0.009 (0.020) (0.015) (0.015) (0.015) (0.016) (0.017) Education and training Years of education (z-score) 0.148* 0.147* 0.147* 0.147* 0.147* (0.013) (0.013) (0.013) (0.012) (0.013) 34 Already received training in EICT in the past 0.063 0.057 0.057 0.060 0.060 (0.031) (0.030) (0.029) (0.032) (0.032) Employment and revenues Had a paid work in the last 30d -0.036 -0.037 -0.034 -0.033 (0.022) (0.024) (0.023) (0.023) Worked in EICT in the last 30d 0.038 0.043 0.042 0.041 (0.033) (0.034) (0.032) (0.032) Network Network size (z-score) 0.031** 0.029** 0.029** (0.011) (0.011) (0.011) Proportion of males in the network (z-score) -0.035** -0.033** -0.033** (0.017) (0.016) (0.016) Role model and support Male role model -0.031 -0.030 Table 5 continued from previous page Chose the ICT or the energy vocational training (1) (2) (3) (4) (5) (6) (0.022) (0.022) Female role model 0.060 0.059 (0.040) (0.037) Can seek professional advice from individuals outside 0.042 0.041 the family (0.029) (0.028) Gender attitudes and agency Gender attitudes (z-score) -0.005 (0.010) Agency: input in productive decisions (z-score) -0.009 35 (0.014) Observations 1400 1400 1400 1400 1400 1400 Cohort FE YES YES YES YES YES YES City FE YES YES YES YES YES YES N otes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 . Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrap to compute p-values. For each column, we include cohort and city fixed-effects. The acronym EICT stands for Energy or Information Communication and Technol- ogy. 6 Conclusion This paper provides an overview of the drivers behind vocational training choices in high-paying sectors, with a particular focus on gender differences. Our analysis shows the importance of education and prior training. Education serves as a prerequisite for accessing vocational training, equipping individuals with the necessary knowledge and skills. Moreover, education and exposure to training have the potential to transform their self- perceptions and attitudes. Enhancing educational opportunities can empower youth to challenge traditional gender roles and seek training in high-paying sectors such as EICT. Our results also suggest that women with lower levels of agency may not be able to opt for such training. Additionally, there is some evidence indicating that specific attitudes toward gender roles can restrict women’s entry into high-paying, male-dominated sectors like EICT. Women who perceive their primary role as being caretakers of domestic tasks are less likely to opt for training in these fields. Addressing social norms and promoting changes in gender attitudes are crucial for encouraging women to enter these lucrative sectors. Furthermore, the study emphasizes the importance of social networks and role models. Women with male role models are more likely to choose training in EICT. Surrounding oneself with the right people provides support, guidance, and potential endorsements, possibly increasing women’s confidence and opportunities in these sectors. These findings provide key insights for policies seeking to improve women’s access to training in higher-paying occupations. However, it is important to acknowledge the study’s limitations. First, the findings presented are based on correlations, and no causal relationships can be inferred. Secondly, it is important to recognize that individuals who have not chosen vocational training may still aspire to work in lucrative sectors. Additionally, the vocational training option in the context of the PRO-Jeunes project may be perceived as more attractive due to the high quality of the training and the technical partnerships with recognized industry leaders. Further research is needed to understand additional mechanisms that influence women’s training choices, such as women’s expectations of discrimination in high-paying sectors (often male-dominated) and the types of discrimination they fear. Exploring whether support from men primarily offers relevant advice or serves as a mechanism for women to gain credibility and validation would also be valuable. To conclude, this study sheds light on the factors influencing young women’s and men’s demand for vocational training in the energy and ICT sectors, two high-paying male-dominated sectors. It emphasizes the importance of addressing conservative attitudes toward gender roles, promoting education and training opportunities, and fostering a supportive environment through role models, 36 among other strategies. Future research should continue to explore these topics to develop further effective interventions that create equal opportunities for women in lucrative sectors. 37 References Agodini, R., Uhl, S., and Novak, T. (2004). Factors That Influence Participation in Secondary Voca- tional Education. MPR Reference No. 8879-400. Technical report, Mathematica Policy Research, Inc. Publication Title: Mathematica Policy Research, Inc. ERIC Number: ED518742. Alfonsi, L., Bandiera, O., Bassi, V., Burgess, R., Rasul, I., Sulaiman, M., and Vitali, A. (2020). Tackling youth unemployment: Evidence from a labor market experiment in uganda. Economet- rica, 88(6):2369–2414. Alibhai, S., Buehren, N., Papineni, S., and Pierotti, R. (2017). Crossovers–Female Entrepreneurs Who Enter Male Sectors: Evidence from Ethiopia. Working Paper, World Bank, Washington, DC. Angrist, J. D. and Krueger, A. B. (1991). Does compulsory school attendance affect schooling and earnings? The Quarterly Journal of Economics, 106(4):979–1014. Arias, O., Evans, D. K., and Santos, I. (2019). The skills balancing act in sub-saharan africa: Investing in skills for productivity, inclusivity, and adaptability. Book, World Bank, Washington, DC. Accepted: 2019-05-17T16:43:52Z ISBN: 9781464811494. Ayub, H. (2017). Parental Influence and Attitude of Students towards Technical Education and Vocational Training. International Journal of Information and Education Technology, 7(7):534– 538. Bardasi, E., Sabarwal, S., and Terrell, K. (2011). How do female entrepreneurs perform? Evidence from three developing regions. Small Business Economics, 37(4):417–441. Publisher: Springer. Bench, S. W., Lench, H. C., Liew, J., Miner, K., and Flores, S. A. (2015). Gender gaps in overesti- mation of math performance. Sex roles, 72:536–546. British Council (2020). Developing skills programming through a gender lens. Book, British Council, Washington, DC. Buehren, N. and Van Salisbury, T. (2017). Female Enrollment in Male-Dominated Vocational Training Courses: Preferences and Prospects. Brief, World Bank, Washington, DC. Accepted: 2017-06-30T14:16:28Z. Cameron, A. C., Gelbach, J. B., and Miller, D. L. (2008). Bootstrap-based improvements for inference with clustered errors. The review of economics and statistics, 90(3):414–427. Campos, F., Frese, M., Goldstein, M., Iacovone, L., Johnson, H., McKenzie, D., and Mensmann, M. (2018). Personal Initiative Training Leads to Remarkable Growth of Women-Owned Small Businesses in Togo. Brief, World Bank, Washington, DC. Accepted: 2018-01-18T15:26:31Z. Campos, F., Goldstein, M., McGorman, L., Munoz Boudet, A. M., and Pimhidzai, O. (2015). Breaking the Metal Ceiling: Female Entrepreneurs Who Succeed in Male-Dominated Sectors. Working Paper, World Bank, Washington, DC. Accepted: 2015-12-18T22:30:21Z. Cassirer, N. and Addati (2007). Expanding women’s employment opportunities: Informal economy workers and the need for childcare. 38 ote d’ivoire jobs diagnostic: employment, productivity, Christiaensen, L. and Premand, P. (2017). Cˆ and inclusion for poverty reduction. Clark, S., Kabiru, C. W., Laszlo, S., and Muthuri, S. (2019). The impact of childcare on poor urban women’s economic empowerment in africa. Demography, 56(4):1247–1272. Croke, K., Goldstein, M., and Holla, A. (2017). Can Job Training Decrease Women’s Self-Defeating Biases?: Experimental Evidence from Nigeria. Working Paper, World Bank, Washington, DC. Accepted: 2017-07-19T18:08:35Z. Das, S., Delavallade, C., Fashogbon, A., Olatunji Ogunleye, W., and Papineni, S. (2023). Occu- pational sex segregation in agriculture: Evidence on gender norms and socio-emotional skills in nigeria. Agricultural Economics, 54(2):179–219. Das, S. and Kotikula, A. (2019). Gender-Based Employment Segregation: Understanding Causes and Policy Interventions. Working Paper, World Bank, Washington, DC. Accepted: 2019-04- 08T17:21:11Z. Doku, D. T. and Asante, K. O. (2015). Women’s approval of domestic physical violence against wives: analysis of the ghana demographic and health survey. BMC women’s health, 15:1–8. Elu, J. (2018). Gender and Science Education in Sub-Saharan Africa. Journal of African Develop- ment, 20(2):105–110. Publisher: Duke University Press. Gassier, M., Rouanet, L., and Traore, L. (2022). Addressing Gender-Based Segregation through Information: Evidence from a Randomized Experiment in the Republic of Congo. Policy Research Working Papers. The World Bank. Goldstein, M., Gonzalez Martinez, P., and Papineni, S. (2019). Tackling the Global Profitarchy: Gender and the Choice of Business Sector. Working Paper, World Bank, Washington, DC. Accepted: 2019-05-30T20:00:32Z. Heckman, J. J., Stixrud, J., and Urzua, S. (2006). The effects of cognitive and noncognitive abilities on labor market outcomes and social behavior. Journal of Labor economics, 24(3):411–482. ete sur le niveau de vie des m´ Institut National de la Statistique (2015). Enquˆ enages (env) 2015. ref. civ 2015 env v01 m. ote d’ivoire. techreport, International International Labour Organization (2019). State of skills : Cˆ Labour Organization, Washington, DC. Jonathan, W., Anu, M., Kweilin, E., Eric, L., Sandrine, D., Eric, K., James, M., Richard, D., and Mekala, K. (2015). How advancing women’s equality can add $12 trillion to global growth | McKinsey. Kuepie, M., Nordman, C. J., and Roubaud, F. (2009). Education and earnings in urban west africa. Journal of comparative economics, 37(3):491–515. Lafortune, J., Riutort, J., and Tessada, J. (2018). Role Models or Individual Consulting: The Impact of Personalizing Micro-entrepreneurship Training. American Economic Journal: Applied Economics, 10(4):222–245. 39 Psacharopoulos, G. and Patrinos, H. A. (2018). Returns to investment in education: a decennial review of the global literature. Education Economics, 26(5):445–458. Reuben, E., Sapienza, P., and Zingales, L. (2014). How stereotypes impair women’s careers in science. Proceedings of the National Academy of Sciences, 111(12):4403–4408. Reuben, E., Wiswall, M., and Zafar, B. (2017). Preferences and biases in educational choices and labour market expectations: Shrinking the black box of gender. The Economic Journal, 127(604):2153–2186. Sansone, D. (2017). Why does teacher gender matter? Economics of Education Review, 61:9–18. on, P., and Ramirez-Espinoza, F. (2023). Reinforcing the stem pipeline in Sevilla, M. P., Bord´ vocational-technical high schools: The effect of female teachers. Economics of Education Review, 95:102428. The African Development Bank Group (2021). The african development bank group gender strategy and action plan 2021 -2025. techreport, The African Development Bank Group. UNESCO (2012). Women and tvet: report of the unesco-unevoc online conference. UNESCO (2017). Cracking the code: girls’ and women’s education in science, technology, engineer- ing and mathematics (STEM) - UNESCO Biblioth` eque Num´ erique. UNESCO (2022). Transforming technical and vocational education and training for successful and just transitions: UNESCO strategy 2022-2029. ISBN: 9789231005527 Publisher: UNESCO. ote d’ivoire. WAEMU Commission (2018). Harmonized survey on households living standards, cˆ Wang, M.-T. and Degol, J. L. (2017). Gender gap in science, technology, engineering, and mathemat- ics (stem): Current knowledge, implications for practice, policy, and future directions. Educational psychology review, 29:119–140. World Bank (2022). Breaking Barriers: Female Entrepreneurs Who Cross Over to Male-Dominated Sectors. Technical report, World Bank, Washington, DC. Accepted: 2022-02-09T16:04:15Z. Yount, K. M. and Li, L. (2009). Women’s “justification” of domestic violence in egypt. Journal of Marriage and Family, 71(5):1125–1140. 40 A Appendix Figure 9: Robustness check for the classification of male-dominated sectors N otes: This figure shows the percentage of sectors classified as male-dominated sectors. Our reference threshold is 75%. In other words, using the reference threshold, any sector that is at least 75% male is considered a male-dominated sector. 41 Table A.1: Robustness check for the classification of male-dominated sectors Male Female VARIABLES T 75 T 65 T 60 T 75 T 65 T 60 Had a training in MDSs -0.006 0.000 0.000 0.206** 0.179** 0.179** 0.017 0.022 0.022 0.046 0.050 0.050 Worked in MDSs -0.070** -0.070** -0.070** 0.006 0.006 0.006 0.010 0.010 0.010 0.030 0.030 0.030 Proportion of contacts in the network working in MDSs -0.122*** -0.122*** -0.122*** -0.063 -0.063 -0.063 0.018 0.018 0.018 0.063 0.063 0.063 42 Has a role model working in MDSs -0.004 -0.004 -0.004 0.097** 0.097** 0.097** 0.019 0.019 0.019 0.022 0.022 0.022 N otes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 This table shows the regression of training choices on different variables. The outcome is a dummy equal to one when the respondent chose the ICT or energy vocational training (MDSs). For each variable (i.e. row)/gender combination, a separate regression is done. In each column, we vary the threshold used to determine whether a sector is classified as male-dominated. Our reference threshold is T 75, i.e. 75%. In other words, using the reference threshold, any sector that is at least 75% male is considered male-dominated. Note that for each regression, we use cohort and city fixed effects. In addition, we include sociodemographic controls (respondent age, marital status, number of dependent children, proportion of female adults in the household, and household wealth index). Standard errors are clustered by zone. Due to a low number of clusters (6), we rely on wild bootstrap to compute p-values. Table A.2: Survey questions: agency, gender attitudes and attitudes towards domestic violence Measurement Survey question Possible answers When a decision is made in the household to personally pursue an entrepreneurial activity, 1. Myself who usually makes the decision? 2. My spouse When a decision is made in the household about your business, 3. My spouse and myself Agency who usually makes the decision? 4. My mother When a decision is made in the household for the purchase of durable goods, or exceptional household expenses, 5 . My father who usually makes the decision? 6. My parents When a decision is made in the household for minor household (my father and mother) 43 expenses (daily food, etc.), who usually makes the decision? / a family member When a decision is made in the household to manage 7. My parents a serious health issue affecting you, /family member who usually makes the decision? and myself When a decision is made in the household to manage a serious health 8. My in-laws issue affecting you, who usually makes the decision? 9. My in-laws and myself When a decision is made in the household about what tasks you will perform on a given day, who usually makes the decision? 10. Other (specify) By nature, men and women are good at different things. 1. Strongly disagree At work, men can handle difficult conditions better than women. 2. Disagree Gender attitudes 3. Agree 4. Strongly agree Measurement Survey question Possible answers The most important role of women is to take care of the home and to cook. Household expenses are the responsibility of the husband, although his wife may help him. In your opinion, is it justified for a husband to hit or beat his wife if she burns the food? In your opinion, is it right for a husband to hit or beat his wife if she argues with him? 1. Yes Attitudes towards domestic violence In your opinion, is it right for a husband to hit or beat his 2. No wife if she goes out without telling him? 44 In your opinion, is it right for a husband to hit or beat his wife if she neglects the children? In your opinion, is it right for a husband to hit or beat his wife if she refuses to have sex with him? In your opinion, is it right for a husband to hit or beat his wife if she talks about protecting herself from AIDS? B Appendix B 45 Table B.3: Variable Selection VARIABLES Theory LASSO Male LASSO Female Sociodemographic characteristics Age of the respondent (z-score) 1 0 1 Number of dependent children (z-score) 1 0 0 Household wealth index by cohort (z-score) 0 0 1 Education and training Years of education (z-score) 1 1 1 Any training 0 0 0 Already received training in EICT in the past 1 0 0 Employment and revenues Wage-employed in the last 30d 0 0 0 Self-employed in the last 30d 0 0 0 Had a paid work in the last 30d 1 0 0 Worked in EICT in the last 30d 1 0 0 Revenues earned in the last 30d (z-score) 0 0 0 Network Has a professional Network 0 0 1 Network size (z-score) 1 0 0 Proportion of males in the network (z-score) 1 0 0 Role model and support Male role model 1 0 1 Female role model 1 0 0 Role model in EICT 0 0 0 Can ask for professional advice from people around 1 0 1 him/her (outside the family) Gender roles and agency Gender attitudes (z-score) 1 0 0 Agrees that women’s most important role is to cook 0 0 0 and take care of her househol Agrees that household expenses are the responsibil- 0 0 0 ity of the husband Agrees that by nature men and women have different 0 0 0 abilities in differenta areas Agrees that at work, men cope better with difficult 0 0 0 conditions than women 46 Table B.3 continued from previous page VARIABLES Theory LASSO MALE LASSO FEMALE Agency: input in productive decisions (z-score) 1 0 0 Attitude towards domestic violence (z-score) 0 0 0 N otes: This table shows the variables selected for table 4 and table 5 (column 1), as well as the variables selected by a LASSO when the sample is restricted to men (column 2) and then women (column 3). 47 C Appendix C 48 Table C.4: Regression of training choice in ICT and energy on isolated factors VARIABLES Male Female Male - Female Sociodemographic characteristics Age of the respondent (z-score) -0.012 0.006 -0.018 (0.013) (0.016) Number of dependent children (z-score) -0.010 -0.043** 0.033* (0.010) (0.012) Household wealth index by cohort (z-score) 0.031 0.041** -0.010 (0.020) (0.015) Education and training Years of education (z-score) 0.144* 0.149* -0.005 (0.017) (0.013) Any training -0.006 -0.044 0.038 (0.025) (0.039) Already received training in EICT in the past 0.090** 0.217** -0.128** (0.025) (0.043) Employment and revenues Wage-employed in the last 30d -0.049 -0.052** 0.003 (0.033) (0.010) Self-employed in the last 30d -0.052** 0.107** -0.158* (0.020) (0.023) Had a paid work in the last 30d -0.045 0.017 -0.062** (0.028) (0.009) Worked in EICT in the last 30d 0.048 0.041 0.007 (0.036) (0.079) Revenues earned in the last 30d (z-score) -0.021 0.022 -0.043 (0.013) (0.025) Network Has a professional Network 0.020 0.071** -0.051 (0.027) (0.019) Network size (z-score) 0.026* 0.045** -0.019* (0.007) (0.008) Proportion of males in the network (z-score) -0.010 0.023 -0.033** (0.011) (0.012) Role model and support Male role model 0.000 0.191* -0.191* 49 Table C.4 continued from previous page VARIABLES Male Female Male - Female (0.012) (0.035) Female role model 0.092** -0.021 0.113** (0.028) (0.025) Role model in EICT 0.103** -0.002 0.104 (0.029) (0.073) Can ask for professional advice from people around -0.006 0.106* -0.112** him/her (outside the family) (0.043) (0.028) Gender attitudes and agency Gender attitudes (z-score) -0.025* -0.036 0.011 (0.008) (0.034) Agrees that women’s most important role is to cook -0.098** -0.182* 0.084* and take care of her househol (0.027) (0.029) Agrees that household expenses are the responsibil- -0.042* -0.112* 0.070 ity of the husband (0.012) (0.047) Agrees that by nature men and women have different 0.005 0.018 -0.013 abilities in differenta areas (0.026) (0.056) Agrees that at work, men cope better with difficult -0.027 0.017 -0.044 conditions than women (0.031) (0.055) Agency: input in productive decisions (z-score) -0.013 0.013 -0.025 (0.014) (0.009) Attitude towards domestic violence (z-score) 0.003 -0.021 0.025 (0.008) (0.021) Observations 1400 1128 Cohort FE YES YES YES City FE YES YES YES 50 Table C.4 continued from previous page VARIABLES Male Female Male - Female N otes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 . The first two coloumns correspond to coefficient estimates from an OLS regression of training choice on respondent characteristics as specified in Equation 1, separately for men (first column) and women (second column). In the third column, we show the difference between the two coefficients. 51 D Appendix D 52 Table D.5: Characteristics of role models VARIABLES Male Female Role models (female and male) Has a role model 0.78 0.67 Female role model 0.07 0.33 Male role model 0.71 0.34 Role model is a friend 0.21 0.16 Role model is a family member 0.30 0.34 Has a role model working in a Male Dominated Sec- 0.27 0.11 tor Role model in EICT 0.09 0.07 Female role models Role model is a female friend 0.01 0.08 Role model is a female family member 0.04 0.15 Has a female role model working in a Male Domi- 0.01 0.04 nated Sector Has a female role model working in the energy or 0.00 0.05 ICT sectors Male role models Role model is a male friend 0.20 0.07 Role model is a male family member 0.25 0.18 Has a male role model working in a Male Dominated 0.26 0.07 Sector Has a male role model working in the energy or ICT 0.08 0.02 sectors N otes: This table shows the descriptive characteristics of role models by gender. The 2nd and 3rd column show the mean for each gender. 53