The presenters introduced a pilot randomized controlled trial in Ghana that examined soft skills development for enhanced employability among students in Technical and Vocational Education and Training (TVET). They explained that the study was supported by J-PAL/MIT and implemented in partnership with the Ghana TVET system and the Association of Ghana Industries. The presentation was delivered by Richard Kwasi Boso Kwadwo Arhin, Patricia Ampofowaa Boso, and R. Kingsford Otoo.
The presenters explained that technical training alone might not be sufficient to produce graduates who are fully prepared for the workplace. They noted that employers increasingly expect TVET graduates to possess not only technical and trade-specific competencies but also a range of soft skills, including communication, teamwork, problem-solving, punctuality, and professional behaviour. These competencies were reportedly identified consistently as areas of weakness among graduates.
They further noted that although TVET institutions were expected to develop these employability competencies, the approaches used to deliver soft-skills training varied considerably across schools. Consequently, the central policy issue was not whether soft skills were important, but rather which delivery model was more effective, feasible, and scalable within public TVET systems.
The presenters described two scalable models for strengthening students' employability skills: Workplace Experiential Learning (WEL) and Project-Based Learning (PBL). WEL was designed to take place in actual workplace settings, where students could gain experience through participation in real work environments, receive feedback from supervisors, and become familiar with professional norms and expectations. PBL, by contrast, was implemented within schools and focused on developing teamwork, problem-solving, communication, leadership, and other employability skills through structured project-based activities.
The control group continued with usual school practice, which consisted of regular TVET instruction without additional structured WEL or PBL interventions. Under the study design, WEL involved four weeks of workplace experiential learning during the vacation period in August and September 2026, while PBL involved four weeks of six structured project-based learning lessons during the semester in June and July 2026.
The presenters argued that both PBL and WEL deserved rigorous evaluation because they represented credible pathways for developing TVET students' soft skills and job readiness. They characterized PBL as a structured, school-based form of experiential learning and WEL as an approach that provided authentic exposure to workplace environments.
They explained that experiential learning theory emphasizes the importance of concrete experience, reflection, conceptualization, and active experimentation. They referred to the work of Kolb (1984) and Dewey (1938) in explaining this theoretical foundation. They also noted that workplace learning is situated within real social and professional environments, allowing learners to develop capabilities through participation in communities of practice, drawing on perspectives associated with Lave and Wenger (1991) and Billett (2001).
According to the presenters, the relevance of this theoretical perspective to the study was that soft skills were fundamentally behavioural capabilities. Such capabilities were more likely to develop when students were provided with opportunities to practice them, receive feedback, reflect on their experiences, adapt their behaviour, and explain what they had learned.
The presenters emphasized that the comparison between PBL and WEL was important because the two models offered different mechanisms and faced different implementation challenges. PBL relied primarily on trained facilitators and structured learning materials, whereas WEL depended on employer placements and workplace mentors. PBL's main strength was its ability to provide relatively standardized opportunities for practice, while WEL's major advantage was its exposure to authentic workplace norms and expectations.
They explained that PBL could potentially feel simulated because learning occurred within the school environment, while the quality of WEL could vary depending on the workplace and the quality of supervision provided by employers. Similarly, scaling PBL could be constrained by instructor quality, whereas scaling WEL could be constrained by the availability and quality of workplace placements.
The presenters therefore argued that both models were credible but that their relative effectiveness, domain-specific effects, and implementation feasibility within Ghana's TVET system remained uncertain. This uncertainty, they suggested, made the comparison a genuine model-selection problem rather than a generic question of whether skills training worked. They also situated the study within broader policy discussions and evidence concerning core work skills, vocational education and training, work-based learning, and skills development, referring to relevant work by the ILO, OECD, European Training Foundation, World Bank, and UNICEF.
The presenters stated that the trial was guided by three principal research questions. First, the study sought to determine the impact of Workplace Experiential Learning on students' soft skills and employability readiness. Second, it examined the impact of Project-Based Learning on students' soft skills and employability readiness. Third, it sought to compare WEL and PBL in terms of their effectiveness, feasibility, and scalability within Ghana's pre-tertiary TVET system.
The presenters explained that the study was embedded within Ghana's public TVET system and focused on Year 2 students enrolled in three-year programmes consisting of two semesters per year. The study involved students from nine technical institutes and covered several trades, including Automotive Technology, Electrical Technology, and Fashion and Design.
They noted that gender participation varied across the trade programmes, with girls accounting for approximately 23–26 percent of participants. The study was implemented within a national TVET context in which pre-tertiary TVET had been provided free of charge since 2017.
For data collection, the researchers used state-provided tablets equipped with KoboCollect, enabling digital collection of study data. The implementation was also supported by vacation industrial attachments, ICT coordinators, and WEL coordinators who facilitated activities within the schools.
The presenters highlighted several practical implementation constraints. In particular, TVET institutions operate within tightly structured 12–16-week semesters, requiring careful coordination of the intervention activities with existing school calendars. The study therefore adopted a structured 4+4+2 week delivery arrangement, taking account of semester schedules and transition periods.
They explained that the study had a broad geographical reach and was connected to Ghana's 16 administrative regions, with implementation involving the Greater Accra, Volta, and Ashanti Regions. Overall, the study was presented as an effort to generate rigorous evidence on how different approaches to soft-skills development could improve the employability of TVET graduates while remaining feasible and scalable within Ghana's public education system.
The presenters explained that the study employed an individual-level randomized controlled trial to enable a fair comparison of three realistic policy options for developing soft skills among TVET students. Randomization was conducted within school–programme blocks, with students assigned to one of three study arms: Workplace Experiential Learning (WEL), Project-Based Learning (PBL), or the control group. A total of 1,842 Year 2 students were selected for the study, of whom 690 were assigned to WEL, 577 to PBL, and 575 to the control group.
They explained that random assignment was important because it allowed the researchers to determine whether differences in outcomes could be attributed to the intervention models rather than to pre-existing differences among students. The primary outcome was a standardized soft-skills index, together with measures of employability and work readiness. The researchers also planned to examine questions concerning for whom the interventions worked, as well as secondary outcomes such as professional confidence, mechanisms of impact, and possible spillover effects.
The presenters explained that WEL and PBL were expected to influence students through different learning mechanisms. WEL was expected to work through exposure to real workplaces, feedback from supervisors, and exposure to professional norms. These experiences were expected to strengthen students' soft-skills confidence and work readiness.
PBL, on the other hand, was expected to operate through structured group projects that required students to collaborate, solve problems, communicate, take leadership roles, and reflect on their experiences. These activities were also expected to improve soft-skills confidence and work readiness.
They summarized the underlying theory of change as a sequence in which workplace exposure or project-based tasks lead to structured practice, which is followed by feedback and reflection, resulting in soft-skills development and ultimately improved work readiness and professional confidence.
The presenters noted that the study used digital field systems to support standardized, transparent, and efficient data collection. Students' tablets were used with the KoboCollect platform, while enumerators conducted group-based data-collection sessions. ICT and WEL coordinators provided implementation support, and routers and backup power systems were used to address connectivity and electricity challenges.
They explained that digital group administration reduced the need for paper-based data collection, helped standardize the administration process, and enabled more effective real-time monitoring of field activities. The digital system was therefore considered an important component of ensuring data quality and transparency throughout the trial.
The researchers explained that randomization was structured around school and trade, with gender blocks used where feasible. Students were allocated to the PBL, WEL, or control arms. The principal outcomes were measures of soft skills and job readiness.
The study was designed to make three key comparisons: PBL versus the control group, WEL versus the control group, and WEL versus PBL. The researchers planned to estimate these effects using an ANCOVA model. In this model, one coefficient would represent the impact of PBL relative to the control group, while another would represent the impact of WEL relative to the control group. The difference between the two intervention coefficients would then provide an estimate of the difference between WEL and PBL.
The presenters emphasized that the primary estimand would be the intent-to-treat (ITT) effect, which measures the impact of being assigned to a particular intervention model. The analysis would preserve the randomization blocks and adjust for baseline outcome levels in order to improve statistical precision.
The presenters emphasized that school calendars were not merely logistical considerations but formed an important part of the intervention environment. They reported that the study had been assessing the feasibility of school-based group administration, the responsiveness of school principals and coordinators, and logistical constraints associated with academic calendars.
They also highlighted several system-readiness considerations, including the reliability of tablets, internet connectivity, and electricity during fieldwork. The feasibility of using non-trade teachers as PBL instructors was also being assessed. For WEL, the researchers were examining the readiness of employers and worksites to provide appropriate placements and learning opportunities for students.
The presenters acknowledged several risks that could affect the validity and implementation of the trial and explained the measures being taken to mitigate them.
One concern was the possibility of spillovers between intervention groups. To address this, the researchers planned to separate intervention schedules, monitor relevant survey items, and conduct checks among control-group participants. Uneven implementation across schools was another concern, and this was being addressed through fidelity tools, session logs, and support from coordinators.
Student attrition was identified as another potential risk. The researchers planned to reduce attrition through group-based data collection and close coordination with schools. They also recognized the possibility of social-desirability bias in students' responses and planned to address this through confidentiality assurances and standardized instructions.
Digital-system failures were another potential challenge. The research team planned to mitigate these through routers, ICT support, backup power, and KoboCollect data-quality checks. Overall, the presenters stressed the importance of transparency in identifying implementation risks and documenting how these risks were managed.
The presenters emphasized that the ultimate purpose of the study was to generate evidence that could inform decisions about whether public TVET systems should invest in WEL, PBL, or a combination of both.
They explained that if WEL produced larger effects, the policy implication would be to strengthen investment in employer-placement systems and workplace learning infrastructure. If PBL produced effects comparable to WEL but at a lower cost, policymakers could consider scaling school-based soft-skills projects as a more cost-effective option.
They also indicated that if the effects varied substantially by gender or trade, policymakers might consider targeting different delivery models to particular student groups or programmes. Similarly, if implementation fidelity proved to be a major determinant of outcomes, investment in coordinators, facilitators, and implementation-support systems would become an important policy priority.
The research team stated that they would particularly value feedback from the African Evidence Summit on several aspects of the study. These included identifying the most appropriate measures of soft skills in African TVET settings, strengthening the measurement of employer-linked workplace experiential learning, and determining how WEL fidelity could be assessed in informal and semi-formal work environments across Sub-Saharan Africa.
They also sought views on how to balance internal validity with policy scalability and which cost-effectiveness measures would be most useful to funders and government ministries when considering investment in soft-skills interventions.
In concluding the presentation, the researchers emphasized that the study was intended to build credible African evidence on which TVET soft-skills delivery model deserved public investment. They characterized the study as rigorous but accessible, African-led but internationally relevant, and methodologically credible while remaining focused on practical policy use.
They explained that the study sought to move beyond the general assertion that soft skills were important and instead generate evidence about which delivery model works, for whom, under what conditions, and at what cost. This evidence would provide policymakers with a stronger basis for decisions concerning the future design and scaling of employability interventions within Ghana's public TVET system.
The presenters identified the research team as including Kwadwo Arhin of USTED, Richard K. Otoo of ACDI, Richard K. Boso of GIMPA, Patricia Ampofowaa of UNISA and GIZ, and Rob Garlick of Duke University. They acknowledged the institutional and research partners supporting the study and noted the importance of collaboration in generating rigorous evidence on TVET and employability in Africa.
The presenters also outlined a potential pathway for expanding the research beyond the initial Ghana pilot. They explained that the Ghana pilot could provide a foundation for a future Ghana–Kenya comparative randomized controlled trial, followed by broader replication and comparative learning across Sub-Saharan Africa. Possible future expansion was identified in countries such as Rwanda, Uganda, and South Africa.
This proposed regional pathway would allow researchers to examine whether the findings from Ghana could be replicated in different institutional, labour-market, and educational contexts and would contribute to a broader African evidence base on effective models for developing employability skills.
In the technical discussion, the researchers explained that the study's minimum detectable effect would be reported in standard-deviation units. The primary powered outcomes would be the soft-skills index and the job-readiness index, while domain-specific effects would be treated as secondary outcomes.
They noted that statistical precision would be strengthened through baseline adjustment, school-by-trade randomization blocks, and gender blocks where these were used. The final reporting would include the number of students randomized to each arm, endline retention, the effective sample after matching, and the estimated minimum detectable effects for the two primary indices.
The researchers cautioned that subgroup analyses by gender, trade, or school should be interpreted carefully and treated as exploratory unless the study was sufficiently powered to make reliable subgroup comparisons. They emphasized that the study would avoid overinterpreting subgroup differences.
The presenters distinguished between the primary ITT analysis and possible secondary analyses involving treatment compliance. They explained that the ITT estimate would measure the effect of being assigned to WEL or PBL and was the most relevant estimate for a policy decision about whether to offer a particular intervention.
They noted that a Local Average Treatment Effect or Treatment-on-the-Treated analysis could be considered if non-compliance were substantial and the first-stage relationship between assignment and participation were sufficiently strong. However, such analyses would be secondary because participation or dosage was not itself randomly assigned.
Attendance and hours of participation could be used as interpretive evidence to help explain outcomes, but the researchers emphasized that dosage should not be interpreted as causal in the same way as randomized assignment. They therefore reiterated that ITT would remain the primary analytical approach because it was directly protected by the study's randomization design.
The researchers explained that the analysis would pre-specify which variables would be used to improve precision, which analyses would be exploratory, and which robustness checks would be conducted to assess the credibility of the findings.
The main covariate adjustment would include the baseline value of the outcome and randomization-block fixed effects. Clean pre-treatment covariates could also be included where they had been pre-specified. The purpose of these adjustments would be to improve precision and address residual imbalance without altering the underlying randomized design.
The team planned exploratory heterogeneity analyses by gender, trade, baseline skill level, school context, and prior workplace or project experience. However, they cautioned that such subgroup patterns would not be treated as strong evidence unless the study had sufficient statistical power.
Robustness checks would include assessments of baseline balance, endline retention and attrition, alternative approaches to constructing outcome indices, complete-case analysis, an 80-percent item-response rule, and spillover checks where relevant. The researchers stressed that these robustness analyses were intended to test the main claims rather than generate an entirely new set of findings.
The presenters explained that the study's outcome measures were directly aligned with its intervention logic. The soft-skills measures included communication, teamwork, problem-solving, adaptability, and professionalism. Job-readiness measures included workplace behaviour, task confidence, professional identity, and labour-market confidence. They noted that the relevant questionnaire sections had been designed to capture these dimensions.
They identified an important evidence gap, arguing that soft-skills delivery is often assumed to be effective without being rigorously tested. They emphasized that PBL and WEL operate through different mechanisms and face different implementation risks, while comparative evidence on TVET delivery models in Ghana and Sub-Saharan Africa remains limited.
The theoretical foundation of the study was linked to the work of Dewey, Kolb, Lave and Wenger, Billett, and Hmelo-Silver. The presenters also referred to literature on employability and TVET from the International Labour Organization, OECD, and European Training Foundation, as well as recent empirical studies examining PBL and work-based learning.
Overall, the presenters concluded that the comparison between PBL and WEL was justified because both models were theoretically credible, highly relevant to policy, and insufficiently tested under real-world Ghanaian TVET conditions. The randomized evaluation was therefore positioned as an important contribution to evidence-based TVET policy in Ghana and potentially across Sub-Saharan Africa.
The presenters explained that technical training alone might not be sufficient to produce graduates who are fully prepared for the workplace. They noted that employers increasingly expect TVET graduates to possess not only technical and trade-specific competencies but also a range of soft skills, including communication, teamwork, problem-solving, punctuality, and professional behaviour. These competencies were reportedly identified consistently as areas of weakness among graduates.
They further noted that although TVET institutions were expected to develop these employability competencies, the approaches used to deliver soft-skills training varied considerably across schools. Consequently, the central policy issue was not whether soft skills were important, but rather which delivery model was more effective, feasible, and scalable within public TVET systems.
The presenters described two scalable models for strengthening students' employability skills: Workplace Experiential Learning (WEL) and Project-Based Learning (PBL). WEL was designed to take place in actual workplace settings, where students could gain experience through participation in real work environments, receive feedback from supervisors, and become familiar with professional norms and expectations. PBL, by contrast, was implemented within schools and focused on developing teamwork, problem-solving, communication, leadership, and other employability skills through structured project-based activities.
The control group continued with usual school practice, which consisted of regular TVET instruction without additional structured WEL or PBL interventions. Under the study design, WEL involved four weeks of workplace experiential learning during the vacation period in August and September 2026, while PBL involved four weeks of six structured project-based learning lessons during the semester in June and July 2026.
The presenters argued that both PBL and WEL deserved rigorous evaluation because they represented credible pathways for developing TVET students' soft skills and job readiness. They characterized PBL as a structured, school-based form of experiential learning and WEL as an approach that provided authentic exposure to workplace environments.
They explained that experiential learning theory emphasizes the importance of concrete experience, reflection, conceptualization, and active experimentation. They referred to the work of Kolb (1984) and Dewey (1938) in explaining this theoretical foundation. They also noted that workplace learning is situated within real social and professional environments, allowing learners to develop capabilities through participation in communities of practice, drawing on perspectives associated with Lave and Wenger (1991) and Billett (2001).
According to the presenters, the relevance of this theoretical perspective to the study was that soft skills were fundamentally behavioural capabilities. Such capabilities were more likely to develop when students were provided with opportunities to practice them, receive feedback, reflect on their experiences, adapt their behaviour, and explain what they had learned.
The presenters emphasized that the comparison between PBL and WEL was important because the two models offered different mechanisms and faced different implementation challenges. PBL relied primarily on trained facilitators and structured learning materials, whereas WEL depended on employer placements and workplace mentors. PBL's main strength was its ability to provide relatively standardized opportunities for practice, while WEL's major advantage was its exposure to authentic workplace norms and expectations.
They explained that PBL could potentially feel simulated because learning occurred within the school environment, while the quality of WEL could vary depending on the workplace and the quality of supervision provided by employers. Similarly, scaling PBL could be constrained by instructor quality, whereas scaling WEL could be constrained by the availability and quality of workplace placements.
The presenters therefore argued that both models were credible but that their relative effectiveness, domain-specific effects, and implementation feasibility within Ghana's TVET system remained uncertain. This uncertainty, they suggested, made the comparison a genuine model-selection problem rather than a generic question of whether skills training worked. They also situated the study within broader policy discussions and evidence concerning core work skills, vocational education and training, work-based learning, and skills development, referring to relevant work by the ILO, OECD, European Training Foundation, World Bank, and UNICEF.
The presenters stated that the trial was guided by three principal research questions. First, the study sought to determine the impact of Workplace Experiential Learning on students' soft skills and employability readiness. Second, it examined the impact of Project-Based Learning on students' soft skills and employability readiness. Third, it sought to compare WEL and PBL in terms of their effectiveness, feasibility, and scalability within Ghana's pre-tertiary TVET system.
The presenters explained that the study was embedded within Ghana's public TVET system and focused on Year 2 students enrolled in three-year programmes consisting of two semesters per year. The study involved students from nine technical institutes and covered several trades, including Automotive Technology, Electrical Technology, and Fashion and Design.
They noted that gender participation varied across the trade programmes, with girls accounting for approximately 23–26 percent of participants. The study was implemented within a national TVET context in which pre-tertiary TVET had been provided free of charge since 2017.
For data collection, the researchers used state-provided tablets equipped with KoboCollect, enabling digital collection of study data. The implementation was also supported by vacation industrial attachments, ICT coordinators, and WEL coordinators who facilitated activities within the schools.
The presenters highlighted several practical implementation constraints. In particular, TVET institutions operate within tightly structured 12–16-week semesters, requiring careful coordination of the intervention activities with existing school calendars. The study therefore adopted a structured 4+4+2 week delivery arrangement, taking account of semester schedules and transition periods.
They explained that the study had a broad geographical reach and was connected to Ghana's 16 administrative regions, with implementation involving the Greater Accra, Volta, and Ashanti Regions. Overall, the study was presented as an effort to generate rigorous evidence on how different approaches to soft-skills development could improve the employability of TVET graduates while remaining feasible and scalable within Ghana's public education system.
The presenters explained that the study employed an individual-level randomized controlled trial to enable a fair comparison of three realistic policy options for developing soft skills among TVET students. Randomization was conducted within school–programme blocks, with students assigned to one of three study arms: Workplace Experiential Learning (WEL), Project-Based Learning (PBL), or the control group. A total of 1,842 Year 2 students were selected for the study, of whom 690 were assigned to WEL, 577 to PBL, and 575 to the control group.
They explained that random assignment was important because it allowed the researchers to determine whether differences in outcomes could be attributed to the intervention models rather than to pre-existing differences among students. The primary outcome was a standardized soft-skills index, together with measures of employability and work readiness. The researchers also planned to examine questions concerning for whom the interventions worked, as well as secondary outcomes such as professional confidence, mechanisms of impact, and possible spillover effects.
The presenters explained that WEL and PBL were expected to influence students through different learning mechanisms. WEL was expected to work through exposure to real workplaces, feedback from supervisors, and exposure to professional norms. These experiences were expected to strengthen students' soft-skills confidence and work readiness.
PBL, on the other hand, was expected to operate through structured group projects that required students to collaborate, solve problems, communicate, take leadership roles, and reflect on their experiences. These activities were also expected to improve soft-skills confidence and work readiness.
They summarized the underlying theory of change as a sequence in which workplace exposure or project-based tasks lead to structured practice, which is followed by feedback and reflection, resulting in soft-skills development and ultimately improved work readiness and professional confidence.
The presenters noted that the study used digital field systems to support standardized, transparent, and efficient data collection. Students' tablets were used with the KoboCollect platform, while enumerators conducted group-based data-collection sessions. ICT and WEL coordinators provided implementation support, and routers and backup power systems were used to address connectivity and electricity challenges.
They explained that digital group administration reduced the need for paper-based data collection, helped standardize the administration process, and enabled more effective real-time monitoring of field activities. The digital system was therefore considered an important component of ensuring data quality and transparency throughout the trial.
The researchers explained that randomization was structured around school and trade, with gender blocks used where feasible. Students were allocated to the PBL, WEL, or control arms. The principal outcomes were measures of soft skills and job readiness.
The study was designed to make three key comparisons: PBL versus the control group, WEL versus the control group, and WEL versus PBL. The researchers planned to estimate these effects using an ANCOVA model. In this model, one coefficient would represent the impact of PBL relative to the control group, while another would represent the impact of WEL relative to the control group. The difference between the two intervention coefficients would then provide an estimate of the difference between WEL and PBL.
The presenters emphasized that the primary estimand would be the intent-to-treat (ITT) effect, which measures the impact of being assigned to a particular intervention model. The analysis would preserve the randomization blocks and adjust for baseline outcome levels in order to improve statistical precision.
The presenters emphasized that school calendars were not merely logistical considerations but formed an important part of the intervention environment. They reported that the study had been assessing the feasibility of school-based group administration, the responsiveness of school principals and coordinators, and logistical constraints associated with academic calendars.
They also highlighted several system-readiness considerations, including the reliability of tablets, internet connectivity, and electricity during fieldwork. The feasibility of using non-trade teachers as PBL instructors was also being assessed. For WEL, the researchers were examining the readiness of employers and worksites to provide appropriate placements and learning opportunities for students.
The presenters acknowledged several risks that could affect the validity and implementation of the trial and explained the measures being taken to mitigate them.
One concern was the possibility of spillovers between intervention groups. To address this, the researchers planned to separate intervention schedules, monitor relevant survey items, and conduct checks among control-group participants. Uneven implementation across schools was another concern, and this was being addressed through fidelity tools, session logs, and support from coordinators.
Student attrition was identified as another potential risk. The researchers planned to reduce attrition through group-based data collection and close coordination with schools. They also recognized the possibility of social-desirability bias in students' responses and planned to address this through confidentiality assurances and standardized instructions.
Digital-system failures were another potential challenge. The research team planned to mitigate these through routers, ICT support, backup power, and KoboCollect data-quality checks. Overall, the presenters stressed the importance of transparency in identifying implementation risks and documenting how these risks were managed.
The presenters emphasized that the ultimate purpose of the study was to generate evidence that could inform decisions about whether public TVET systems should invest in WEL, PBL, or a combination of both.
They explained that if WEL produced larger effects, the policy implication would be to strengthen investment in employer-placement systems and workplace learning infrastructure. If PBL produced effects comparable to WEL but at a lower cost, policymakers could consider scaling school-based soft-skills projects as a more cost-effective option.
They also indicated that if the effects varied substantially by gender or trade, policymakers might consider targeting different delivery models to particular student groups or programmes. Similarly, if implementation fidelity proved to be a major determinant of outcomes, investment in coordinators, facilitators, and implementation-support systems would become an important policy priority.
The research team stated that they would particularly value feedback from the African Evidence Summit on several aspects of the study. These included identifying the most appropriate measures of soft skills in African TVET settings, strengthening the measurement of employer-linked workplace experiential learning, and determining how WEL fidelity could be assessed in informal and semi-formal work environments across Sub-Saharan Africa.
They also sought views on how to balance internal validity with policy scalability and which cost-effectiveness measures would be most useful to funders and government ministries when considering investment in soft-skills interventions.
In concluding the presentation, the researchers emphasized that the study was intended to build credible African evidence on which TVET soft-skills delivery model deserved public investment. They characterized the study as rigorous but accessible, African-led but internationally relevant, and methodologically credible while remaining focused on practical policy use.
They explained that the study sought to move beyond the general assertion that soft skills were important and instead generate evidence about which delivery model works, for whom, under what conditions, and at what cost. This evidence would provide policymakers with a stronger basis for decisions concerning the future design and scaling of employability interventions within Ghana's public TVET system.
The presenters identified the research team as including Kwadwo Arhin of USTED, Richard K. Otoo of ACDI, Richard K. Boso of GIMPA, Patricia Ampofowaa of UNISA and GIZ, and Rob Garlick of Duke University. They acknowledged the institutional and research partners supporting the study and noted the importance of collaboration in generating rigorous evidence on TVET and employability in Africa.
The presenters also outlined a potential pathway for expanding the research beyond the initial Ghana pilot. They explained that the Ghana pilot could provide a foundation for a future Ghana–Kenya comparative randomized controlled trial, followed by broader replication and comparative learning across Sub-Saharan Africa. Possible future expansion was identified in countries such as Rwanda, Uganda, and South Africa.
This proposed regional pathway would allow researchers to examine whether the findings from Ghana could be replicated in different institutional, labour-market, and educational contexts and would contribute to a broader African evidence base on effective models for developing employability skills.
In the technical discussion, the researchers explained that the study's minimum detectable effect would be reported in standard-deviation units. The primary powered outcomes would be the soft-skills index and the job-readiness index, while domain-specific effects would be treated as secondary outcomes.
They noted that statistical precision would be strengthened through baseline adjustment, school-by-trade randomization blocks, and gender blocks where these were used. The final reporting would include the number of students randomized to each arm, endline retention, the effective sample after matching, and the estimated minimum detectable effects for the two primary indices.
The researchers cautioned that subgroup analyses by gender, trade, or school should be interpreted carefully and treated as exploratory unless the study was sufficiently powered to make reliable subgroup comparisons. They emphasized that the study would avoid overinterpreting subgroup differences.
The presenters distinguished between the primary ITT analysis and possible secondary analyses involving treatment compliance. They explained that the ITT estimate would measure the effect of being assigned to WEL or PBL and was the most relevant estimate for a policy decision about whether to offer a particular intervention.
They noted that a Local Average Treatment Effect or Treatment-on-the-Treated analysis could be considered if non-compliance were substantial and the first-stage relationship between assignment and participation were sufficiently strong. However, such analyses would be secondary because participation or dosage was not itself randomly assigned.
Attendance and hours of participation could be used as interpretive evidence to help explain outcomes, but the researchers emphasized that dosage should not be interpreted as causal in the same way as randomized assignment. They therefore reiterated that ITT would remain the primary analytical approach because it was directly protected by the study's randomization design.
The researchers explained that the analysis would pre-specify which variables would be used to improve precision, which analyses would be exploratory, and which robustness checks would be conducted to assess the credibility of the findings.
The main covariate adjustment would include the baseline value of the outcome and randomization-block fixed effects. Clean pre-treatment covariates could also be included where they had been pre-specified. The purpose of these adjustments would be to improve precision and address residual imbalance without altering the underlying randomized design.
The team planned exploratory heterogeneity analyses by gender, trade, baseline skill level, school context, and prior workplace or project experience. However, they cautioned that such subgroup patterns would not be treated as strong evidence unless the study had sufficient statistical power.
Robustness checks would include assessments of baseline balance, endline retention and attrition, alternative approaches to constructing outcome indices, complete-case analysis, an 80-percent item-response rule, and spillover checks where relevant. The researchers stressed that these robustness analyses were intended to test the main claims rather than generate an entirely new set of findings.
The presenters explained that the study's outcome measures were directly aligned with its intervention logic. The soft-skills measures included communication, teamwork, problem-solving, adaptability, and professionalism. Job-readiness measures included workplace behaviour, task confidence, professional identity, and labour-market confidence. They noted that the relevant questionnaire sections had been designed to capture these dimensions.
They identified an important evidence gap, arguing that soft-skills delivery is often assumed to be effective without being rigorously tested. They emphasized that PBL and WEL operate through different mechanisms and face different implementation risks, while comparative evidence on TVET delivery models in Ghana and Sub-Saharan Africa remains limited.
The theoretical foundation of the study was linked to the work of Dewey, Kolb, Lave and Wenger, Billett, and Hmelo-Silver. The presenters also referred to literature on employability and TVET from the International Labour Organization, OECD, and European Training Foundation, as well as recent empirical studies examining PBL and work-based learning.
Overall, the presenters concluded that the comparison between PBL and WEL was justified because both models were theoretically credible, highly relevant to policy, and insufficiently tested under real-world Ghanaian TVET conditions. The randomized evaluation was therefore positioned as an important contribution to evidence-based TVET policy in Ghana and potentially across Sub-Saharan Africa.