Soft Skills

NBER WORKING PAPER SERIES
MAKING ENTREPRENEURS:
RETURNS TO TRAINING YOUTH IN HARD VERSUS SOFT BUSINESS SKILLS
Laura Chioda
David Contreras-Loya
Paul Gertler
Dana Carney
Working Paper 28845

http://www.nber.org/papers/w28845
NATIONAL BUREAU OF ECONOMIC RESEARCH

1050 Massachusetts Avenue
Cambridge, MA 02138
May 2021
We gratefully acknowledge financial support from CEDIL/UKaid, ILO, Jacobs Foundation,

JPAL PPE program, USAID, and Wellspring Philanthropic Fund. We are also grateful to
Fernando Saltiel, David McKenzie, Susana Puerto, as well as seminars and conferences
participants at Berkeley, LSE, World Bank, IADB, RIDGE, and HEC Paris. We are particularly
grateful to Educate! and Meghan Mahoney. Ada Kwan, Vivian Lo, Aisling Scott, Pooja Suri, and
Nicole Perales provided excellent research assistance. We gratefully acknowledge Nathan Fiala's
contributions to early stages of the design and survey instruments. The experiment described in
this paper is registered at the AEA RCT Registry under the code AEARCTR-0002134 and the
Uganda National Council of Science and Technology (UNCST, SS4310). The protocol was
granted IRB approval by the IPA Institutional Review Board (IPA IRB, 9850) and the Mildmay
Uganda Ethics Review (MUREC, 0104 2017). The authors declare that they have no financial or
material interests in the results discussed in this paper. The views expressed herein are those of
the authors and do not necessarily reflect the views of the National Bureau of Economic
Research.
NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.
© 2021 by Laura Chioda, David Contreras-Loya, Paul Gertler, and Dana Carney. All rights
reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit
permission provided that full credit, including © notice, is given to the source.
Making Entrepreneurs: Returns to Training Youth in Hard Versus Soft Business Skills
Laura Chioda, David Contreras-Loya, Paul Gertler, and Dana Carney
NBER Working Paper No. 28845
May 2021
JEL No. J24,L26,O1
ABSTRACT
We study the medium-term impacts of the Skills for Effective Entrepreneurship Development
(SEED) program, an innovative in-residence 3-week mini-MBA program for high school students
modeled after western business school curricula and adapted to the Ugandan context. The
program featured two separate treatments: the hard-skills MBA features a mix of approximately
75% hard skills and 25% soft skills; the soft skills curriculum has the reverse mix. Using data on
4400 youth from a nationally representative sample in a 3-arm field experiment in Uganda, the
3.5 year follow-up demonstrated that training was effective in improving both hard and soft
skills, but only soft skills were directly linked to improvements in self-efficacy, persuasion, and
negotiation. The skill upgrade was rewarded in substantially higher earnings; 32.1% and 29.8%
increases in earnings for those who attended hard-and soft-training, respectively, most of which,
was generated through self-mployment. Furthermore, youth in both groups were more likely to
start enterprises and more successful in ensuring their businesses' survival. The program led to
significantly larger profits (24.2% and 27.2% for hard- and soft- treatment arms respectively) and
larger business capital investments (38.4% and 32.6% for SEED hard and SEED soft,
respectively). Both SEED curricula were very cost-effective; two months worth of the extra
earnings caused by the training alone would exceed the cost of the program. These benefits
abstract from the job- and business-creation benefits of the program, which were substantial:
relative to the control group, SEED entrepreneurs created 985 additional jobs and 550 new
businesses.
Laura Chioda Paul Gertler

Institute for Business and Social Impact Haas School of Business
Haas School of Business University of California, Berkeley
University of California Berkeley, CA 94720
Berkeley, CA 94720 and NBER
USA gertler@haas.berkeley.edu
lchioda@berkeley.edu
Dana Carney
David Contreras-Loya Haas School of Business
School of Public Health UC Berkeley
UC Berkeley Berkeley, CA 94720
Berkeley, CA 94705 dcarney@berkeley.edu
davidcon@berkeley.edu
1 Introduction
Many commentators have recognized entrepreneurship as a powerful engine of economic growth
(e.g., Smith (1776). Entrepreneurs make investments in risky ventures in hopes of high returns,
introduce new innovative goods and services, and improve production processes (e.g., Schumpeter,
1911). Accordingly, governments have invested heavily in educational programs to increase their
“stock” of entrepreneurial skills (Deming, 2017; OECD, 2015); They spend more than US$ 1 billion
annually on entrepreneurship training (McKenzie et al., 2020).
Unlocking the “production function” of high-quality entrepreneurs is pivotal to addressing nu-
merous development challenges. Small and medium enterprises (SMEs) account for the majority
of businesses worldwide and are important contributors to job creation and global economic devel-
opment; They represent about 90% of businesses and account for more than 50% of employment
worldwide. Formal SMEs generate most formal jobs, and contribute up to 40% of national income
in emerging economies (World Bank).1 SMEs form the backbone of the African economy, repre-
senting more than 90% of businesses and employing about 60%-70% of workers, many of whom are
women and youth (International Trade Centre, 2018). The World Economic Forum estimates that
Africa’s workforce will increase by a staggering 910 million people by 2050, of which 830 million will
be in Sub-Saharan Africa, creating enormous pressure for jobs on SMEs, which typically account
for approximately 80% of new jobs. The stakes to produce high quality entrepreneurs are even
higher considering that the overwhelming majority of business owners are entrepreneurs who lack
alternative opportunities (Schoar, 2010).
To assess the extent to which successful entrepreneurs can be ”made” or trained, we designed
and implemented an at-scale intervention called Skills for Effective Entrepreneurship Development
(SEED) in Uganda. SEED is an innovative, full time, in-residence 3-week mini-MBA program
modeled after western business school curricula, adapted to the Ugandan context.
SEED targets youth at the end of secondary school, which is earlier in life than traditional
entrepreneurial training interventions and before youth begin their working lives—an age group not
yet extensively studied for this type of intervention. This phase represents a unique opportunity to
shape their futures and set them on higher trajectories of lifetime productivity. During adolescence
until early to mid 20s high-level executive and social brain processes needed for emotion regulation,
working memory, inhibitory control, abstract thought, decision making, perspective taking, and
interpersonal communication undergo substantial development (e.g., Blakemore and Robbins, 2012;
Dumontheil, 2014; Sebastian et al., 2010; Andrews et al., 2021; Choudhury et al., 2006).2
SEED was designed to generate transformational entrepreneurial skills. Recent evidence (Schoar,
2010) suggests that transformational and subsistence entrepreneurs are quite different: not only
in terms of their economic objectives, but also with respect to ability, motivation, and person-
1
The figure is significantly higher if informal SMEs are included.
2
Development of these cognitive processes appears to be underpinned by structural and functional brain plasticity.
During adolescence until young adulthood, the prefrontal cortex undergoes significant structural changes that entail
the proliferation of new synapses and rewiring of existing connections particularly between prefrontal and limbic
regions (Ahmed et al., 2015; Blakemore, 2012).
1
ality characteristics. Transformational entrepreneurs tend to have higher ability (de Mel et al.,
2013), stronger confidence/ambition/motivation, and higher expectations for their own business
(enterprise-self efficacy) (Ardagna and Lusardi, 2010). A number of other studies have emphasized
the role of various intra- and inter-personal skills in entrepreneurial success (Lucas, 1978; Baumol,
1996; Murphy et al., 1991; Gennaioli et al., 2013; Levine and Rubinstein, 2017).3
SEED featured two separate treatments arms. The hard skills SEED placed a stronger emphasis
on hard skills: 75% hard skills (e.g., accounting, finance, economics, marketing, etc.) and 25% soft
skills (e.g., such as negotiation, persuasion, grit, emotional regulation, etc.). The second treatment
arm, soft skills SEED, had the reverse mix: 25% hard skills and 75% soft skills. In 2012, a sample
of 4,400 high school students were recruited from a random sample of 200 high schools from the
universe of full-time high schools in Uganda. Study participants were then randomly assigned to
one of the two treatments or to a control group.
Immediately following the intervention, both curricula yielded significant differential accrual of
hard and soft skills knowledge consistent their specialized focus, thereby validating the intent of
the training. During the 3.5 year follow-up, SEED graduates still outperform their counterparts
in the control group in terms of hard skills and business knowledge, as well as in socio-emotional
skills/traits. While test scores of hard skills knowledge for the two treatment groups converged,
the treatment effects on inter- and intra-personal skills tell a more nuanced story. The differential
between hard and soft skill treatments disappeared for intra-personal skills (Stability, Plasticity,
Stress, etc.), while a meaningful difference persisted for inter-personal skills (Negation and Per-
suasion) and Self-Efficacy (defined as the belief about one’s ability to effectively perform tasks,
including interpersonal tasks, needed to achieve a valued goal), favoring the SEED soft treatment.
To put the results that follow into perspective, it is worth noting that the medium term ef-
fects provide insight on a nationally representative sample of young adults (34% female) in their
mid-twenties. 3.5 years post-high school and post-intervention, 72% of the sample was either
self-employed or working for someone else, with nearly 90% of the earnings generated by self-
employment. Study participants accounted for 3,369 businesses created over the course of 3.5
years. We collected detailed economic data on over 1,665 young enterprises (34.7% of which are
formal) that were operating at endline. These enterprises were responsible for creating over 6,500
jobs (6,657 employees including the owners), corresponding to approximately 3.93 employees per
business.
Against this backdrop, our main results indicate that SEED was effective not only at jump-
starting youth human capital (soft and hard skills) accumulation, but also at setting them on
higher earnings trajectories and expanding their income opportunities. Youth in both training
groups are more economically active (SEED Hard: 9.8%; SEED Soft: 7.1%) and earn more (SEED
Hard: 32.1%; SEED Soft: 29.8%) than the control group, with these differences being almost
entirely driven by higher degrees of entrepreneurship. SEED graduates are more likely to start
3
These same skills have been shown to also be important in the context of more general labor market outcomes
(e.g., Borghans et al., 2014; Deming, 2017; Deming and Kahn, 2018; Heckman et al., 2006; Bowles et al., 2001; Autor,
2015).
2
businesses (SEED Hard: 10.5 pp; SEED Soft: 11.1 pp relative to the control group) that survive
over time. Moreover, the program contributed to the creation of vibrant and more profitable
ventures (SEED Hard: 24.2%; SEED Soft:27.2 % higher profits than the control group), with
SEED graduates behaving more like transformational/opportunity entrepreneurs (as in Schoar,
2010). They were more successful in seeding dynamic businesses that grow beyond an individual’s
subsistence needs, leading to expanded employment opportunities for others (985 additional jobs
and 550 new businesses), and attracting larger capital investments (SEED Hard: 42.9%; SEED
Soft: 36%). Both SEED curricula were highly cost-effective and compare favorably to existing
training alternatives: two months worth of earnings alone would exceed the cost of the program
($118 per participant). These benefits abstract from the job- and business-creation benefits of the
program, which were substantial.
Our work stands in contrast to the large body of work has studied the effectiveness of business-
training courses in improving the performance of managers of existing enterprises. Despite their
popularity, the evidence base for training programs has been disappointing. The reviews of
Blattman and Ralston (2015), McKenzie and Woodruff (2017), and Card et al. (2018) document
relatively weak or temporary impacts of training programs in low-income settings. A more recent
meta-analysis (McKenzie, 2020) concludes that traditional entrepreneurship training has modest
but (across studies) significant effects in terms of improving business practices and business out-
comes for microenterprises. The most compelling evidence relates to the training of small-scale
subsistence enterprises: successful training interventions may lead to short-lived improvements in
profits or sales, without altering the growth paths of enterprises.
Fewer resources have been devoted to assessing whether individuals can be taught a more
entrepreneurial personality (Kerr et al., 2017). A small but growing number of attempts have
been made to evaluate interventions that teach such attributes to entrepreneurs in developing
countries. Programs that foster an entrepreneurial mindset have shown more encouraging results
than traditional training. Innovations include personal initiative training (see Campos et al., 2017;
Alibhai et al., 2019; Ubfal et al., 2019, in Togo, Ethiopia, and Jamaica, respectively) and psychology-
influenced training curricula, including interpersonal skills (Dammert and Nansamba, 2019). Efforts
to follow up on Campos et al. (2017)’s success in Togo – where training personal initiative increased
firm profits by 30%, compared to a statistically insignificant 11% increase among those who received
traditional training– yielded largely insignificant and therefore less encouraging results in other
contexts.
Our work also contributes to a broad literature that underscores the importance of youth
interventions and their potential to endow youth with critical intra- and inter-personal skills. This
critical developmental phase has garnered increasing attention (Kautz et al., 2014). Programs
targeting the development of youth’s socio-emotional skills have led to improvements in education
outcomes (Yeager et al., 2019; Alan and Ertac, 2018; Alan et al., 2019) and have shown promise as
effective prevention strategies against crime and antisocial behavior (Blattman et al., 2017; Heller
et al., 2017).
3
Recent work has also highlighted the role of negotiation and persuasion skills – skills that are
at the center of the SEED intervention – in shaping education outcomes and gender attitudes of
adolescent girls. In Zambia, Ashraf et al. (2020) finds that negotiation training (which taught girls
skills to reconcile different interests by looking for “win-win” solutions) led to improvements in
human capital outcomes 3 years after the intervention. In India, Dhar, Jain and Jayachandran
(2020) document meaningful impacts on gender attitudes of a school-based persuasion intervention
for adolescents among the sample of both boys and girls two years after training. No labor market
outcomes were reported by either study.
This paper also relates to Bandiera et al. (2020) and Alfonsi et al. (2020) because of their
focus on youth and on labor market outcomes, although these studies place less emphasis on socio-
emotional skills. Bandiera et al. (2020) assess the impact of a program that provides adolescent
girls with an opportunity to simultaneously acquire (i) vocational skills to enable them start small-
scale income-generating activities and (2) life skills to help them make informed choices about sex,
reproduction, and marriage. Four years post-intervention, adolescent girls in treated communities
are more likely to be self-employed and to delay family formation. Alfonsi et al. (2020) design a labor
market field experiment to compare demand-side (vocational training, VT) and supply-side policies
(offering firms wage subsidies to train workers on-the-job, FT) to tackle youth (approximately 20
years old) unemployment. Their results speak to the importance of portability and transferability
of skills. Indeed, both treatments led to skill accumulation, VT workers learn sector-specific skills,
FT workers learn more firm-specific skills, which translates into differential gains in employment
rates (VT: 21% vs FT: 14%) and increases in total earnings (VT: 34% vs FT: 20%).
The paper is organized as follows. Section 2 details the intervention and its implementation.
Section 3 describes the research design. Section 4 presents the experimental design, data and
estimation strategy. Section 5 presents the impacts of the hard and soft skill treatments on post-
intervention knowledge assessments and on various outcomes in the medium run (3.5 years). It also
undertakes a cost-benefit analysis. Section 6 concludes by highlighting the broader implications of
our findings for policies and future research.
2 The Intervention
The study team designed two separate curricula: a hard skills-focused mini-MBA and a soft skills-
focused mini-MBA. The hard skills curriculum is approximately 75% hard and 25% soft skills
and the soft skills curriculum is the reverse mix. Both are intensive 3-week long entrepreneurship
courses. An outline of the lesson plan for the two curricula is summarized in Figure 1.
Students in the hard skills program differentially focus on (1) accounting, covering financial
statements (income statements and balance sheets) and the cost structure of an enterprise, (2)
business creation covering generating business ideas, selecting a suitable market, resources needed
to start an enterprise, and legal forms of business ownership (3) management, covering sales, keeping
track of business operations, hiring and managing people, and selecting suppliers; and (4) finance,
4
Figure 1: Lesson Plans for Hard and Soft Curricula
covering managing money, investments, and obtaining financing to start a business. This hard-skills
curriculum provides foundational skills so that the students know how to generate a business idea,
how to secure capital, how to get the business off the ground, and how to manage it.
Students in the soft skills program differentially focus on inter- and intrapersonal skills.
Specific topics include (1) learning to control one’s self through self-management, goal setting,
risk taking, decision making and personal power;4 (2) persuasion through communication strate-
gies, leadership and team building, group decision making strategies, social networking and public
speaking, (3) negotiation through win-win strategies and personal power. This soft skills curriculum
provided foundational skills so that students have more self-awareness, better ability to regulate
emotions, delay gratification, and make sound decisions.
The business plan curriculum provided to both groups was designed to provide the elements
of a business plan that students could use as a guide to develop successful business plans. Students
were first provided with the goals, structure and elements of a business plan. The course then
went into depth on key elements including market research, human resources, legal and tax issues,
financial planning, fund raising, and marketing. Two classes were devoted to pitching the business
plan and two others to case studies. The business plan curriculum was designed to guide students
in writing their own original business plan, with the author of the winning business plan at each
host school winning a cash prize of 1 million Ugandan Shillings, UGX ($400). One student per site
4
personal power is the ability to ignore the influence of others, to control one’s own outcomes, and to be personally
independent (van Dijke and Poppe, 2006; Galinsky et al., 2008).
5
won, for a total of 10 winners.
The curricula were designed to enable highly participatory classroom environments, addressing
different modalities of learning through a variety of exercises and teaching methods. The soft skills
curriculum was based on a social entrepreneurship curriculum developed and used by Educate!, an
NGO that delivers social entrepreneurship training and mentorship to Ugandan secondary school
students.5 The hard skills curriculum was mainly based on the Know About Business curriculum
developed and used by the International Labor Organization in Africa. Both soft and hard skills
curricula integrated content from the MBA program of the Haas School of Business at UC Berkeley.
The curricula were extensively pilot tested to ground them in the local language and conceptual
environment, to make them relevant and salient to local business institutions, and to ensure that
course content was appropriate and well-understood, and that the teaching methods were interesting
and engaging.
Each of the curricula were broken down into 17 sessions and detailed lesson plans were developed
for each session (see Figure 1). Lesson plans included detailed steps for the teacher to follow, framed
by objectives and rationale to encourage a sense of ownership by the teacher. The plans covered
learning objectives, key concepts, case applications, presentation materials and a series of interactive
exercises. The lesson plans were designed for use in an intense Western-Socratic based course in
which the learning process encourages sharing, discovery and reflection while maintaining a similar
classroom environment between curricula and across all training sites.6
The training took place between May 6th and May 24th, 2013 – for a total of 17 instruction
days – at 20 host school training sites. Training took place during the break between Term 1 and
Term 2 of the academic year, when dormitories at host schools would be empty. Students slept
and boarded at the schools, attending two lessons each day: one hard or soft skills lesson in the
morning and one business plan lesson in the afternoon. Each student also had two “prep” periods
a day; one in the morning and one in the afternoon. Students were expected to work on homework
or on their business plans during these periods, which were supervised by SEED camp directors
and teachers.
The study used a number of different types of monitoring and supervision to assess both oper-
ational and academic aspects of the training and to ensure that the training was implemented as
designed. The monitoring was used to identify deviations from the curriculum and weak instruc-
tional pedagogy by specific teachers and correct problems in real time.
5
Educate! tackles youth unemployment by partnering with youth, schools and governments to design and deliver
solutions that equip young people in Africa with the needed skills. Educate! advises national governments (Uganda,
Kenya, Rwanda) on curriculum design and teacher training initiatives. Educate! was recognized by the World Bank
(S4YE) and the UN, received the 2015 WISE award and the 2018 Klaus J. Jacobs Prize.
6
Finally, there was also a series of planned social activities that were implemented consistently across all training
sites. These took place nearly every evening and included debates, a cultural night (in which students representing
different cultures rehearsed and performed dances), movie nights, a spelling bee, a lip-syncing competition (“miming”),
stand-up comedy, quiz nights, and a social dance. In addition, sports activities took place each afternoon before
dinner. Basic sports equipment (volleyballs, netballs, and soccer balls) was purchased for each school.
6
3 Experimental Design
The study is designed to be nationally representative, with both students and teachers randomly
assigned to one of three arms (hard skills, soft skills and control) and to one of 20 training sites.
The design of the field experiment is summarized in Figure 2.
3.1 Student Recruitment

To ensure representativeness at the national level, students were recruited from 200 full-time sec-
ondary schools randomly selected from the universe of Uganda’s approximately 700 full-time sec-
ondary schools. In October of 2012, promoters went to each of the 200 schools to recruit students
in their sixth (last) year of secondary school. Each recruiting visit consisted of an oral presentation
and promotional posters. The promotion informed the students of a 3-week in-residence program
designed to train students in entrepreneurship skills. SEED would be free, including lectures, room
and board, and transport, and students would have the chance to write business plans and enter
a competition for a large cash prize. The promoters explained that there was limited space in the
program, such that not all applicants would be admitted, but that they all have an equal chance
of admission through a fair lottery.
Students interested in the program were asked to fill out both a simple application form and
a baseline survey.7 Response to the SEED program opportunity was overwhelmingly positive. In
almost all schools, a majority of the S6 class applied for the program. In fact, 8,080 students
applied. However, 313 (3.9%) were dropped due to problems with their student ID, 113 (1.3%)
because the student failed to indicate their gender, and another 223 were dropped because they
were already participating in another entrepreneurship training program. The remaining 7,431
served as our population universe for the study.
3.2 Student Assignment to Treatment Arms and Training Sites

Power calculations indicated that we needed 1,200 students in each arm (hard skills treatment, soft
skills treatment, and control group), or 30 students in each of the 80 classes. Taking into account
estimates of attrition rates from similar programs and the long period of time between recruitment
and program implementation, the evaluation team over-enrolled each treatment arm by one third,
accepting approximately 1,600 students in each arm, or 40 students per class. The control group
remained at about 1,200 students. Therefore, we randomly sampled 4,400 students from the 7,436
valid applicant pool to participate in the study, randomly assigning 1,600 to each treatment group
and 1,200 to the control group, stratified by school and gender.
In order to minimize travel time, recruitment schools were clustered into four geographic regions,
with students participating in the study being sent to a randomly selected school in their region.
7
The application form asked students to provide detailed contact information. It also included the question:
“If you participate in the SEED training, what business would you start and how would you start and run it?”
This question was included to indicate to students that SEED would accept serious applicants who have given some
thought to the idea of starting their own businesses.
7
Figure 2: Experimental Design
Of the 200 schools visited during data collection and recruitment, 20 eventually served as training
sites. The distribution of training sites was chosen based on the number of individuals in the sample
surveyed in each region in order to achieve fairly even numbers across regions. Roughly 40% of the
sample attended schools in the West, 40% from the East (including 20% in the Jinja area and 20%
in the Mbale area), and 20% from the North. As a result, eight sites were located in the West,
eight in the East, and four in the North for a total of 20 training sites.
The 20 schools were divided into 10 sites that included one hard skills only school and one soft
skill only school. Each school had two classrooms for a total of 40 classrooms, 20 in each arm of
the study. Students assigned to the hard skills arm were randomly assigned to one of the hard
skills classrooms within their region, stratified by feeder school and gender. Similarly, students
assigned to the soft skill arm were randomly assigned to one of the soft skills sites within their
region, stratified by feeder school and gender. Roughly 160 students were assigned to each site, 80
per school and were divided into two classes of 40 each.
3.3 Teachers Training and Assignment

Each school was staffed with three teachers: two regular curriculum instructors (hard or soft skills),
who both teach the regular curriculum and one instructor who teaches the business plan curriculum
exclusively. Teachers were randomly assigned to one of the training site and to a classroom stratified
by language ability.
The research team worked with Educate! to recruit, hire, and train instructors in general
8
teaching methods. Trainees were required to pay a fee of (approximately) $134 to enroll in the
training and, upon completion, were issued a formal certificate, which is highly valued in Uganda
as it is perceived as strong signal of quality.
Potential instructors were first trained in general (western) teaching methods. Instruction took
place five mornings per week for four weeks. Recruits who met the training requirements were then
randomly assigned to one of three groups: (1) hard skills, (2) soft skills, and (3) business plan. The
second training phase, lasting three weeks, focused exclusively on their assigned topic (hard, soft
or business plan). The training was designed to establish three levels of learning – the “what”, the
“how”, and the “why” of each lesson – including insight into the rationale for the content, thereby
endowing teachers with a deep understanding as well as a sense of ownership over the material.
Teachers received hands-on practice with their assigned lessons via read-through assemblies, mock
lesson presentations and role playing. After the specialized training, the curriculum development
consultants provided an additional day of coaching to several teachers who were identified as being
relatively weak based on a consensus among trainers. All teachers had to pass a knowledge test
satisfactorily (the same that was administered to students) in order to become SEED instructors.
4 Data, Attrition, and Compliance

A baseline survey was administered to all study participants at the time of recruitment and prior
to assignment to the study arm. The survey was framed as purely research-oriented and with no
bearing on a student’s chances of being accepted. The 45 minutes survey collected a rich set of
baseline characteristics: demographic and family background as well as participants’ education,
ability and intelligence proxies, job experience, personality traits and soft-skills.
Table A.1 reports summary statistics for study participants as well as the results of balance
tests between intervention groups.
At baseline, youth were on average 20 years old, 34% female, with 73% attending a boarding
school and 70% planning on attending university. Study participants became active in the labor
market at an early age (14.5 years old on average) and, by the time, they reached their last year
of secondary school, approximately 40% owned a small business. The majority of youth’s parents
had not completed secondary education (60% of fathers and 73% of mothers) and tended largely to
be employed in low skill, manual jobs. As alluded to in the introduction, in low income countries
nearly half of the labor force runs an enterprise, consistent with our sample (51% of households
owned a business).
Randomization achieved good balance across groups, with p-values associated to joint tests of
whether the 30 covariates predict treatment statuses relative to the control exceeding 0.16. We test
for balance across 30 outcomes, such that some individual coefficients are expected to be significant
by chance. Few systematic differences (only 5 out of 90 comparisons) are observed, and were
quantitatively small.
Skills knowledge assessments were carried out before and at the end of the training. These
9
include an extensive array of concepts at the core of the two curricula.8 The skills assessment was
administered in order to establish whether the two treatment arms endow youth in the hard and
soft skills training with different sets of skills that reflect the intensity of the assigned treatment
arm. As discussed later, both treatments did improve knowledge of hard and soft skills, but youth
in the hard skills arm accrued larger gains in terms of hard skills knowledge relative to those who
attended the soft skills arm; the gain in terms of soft skills were reversed.
In the context of the 3.5 year follow-up survey, we collected an extensive set of outcomes that
can be organized into four groups: (1) SES and demographic outcomes (e.g., education, self-assessed
present and future social standing and wealth standings, time use); (2) labor market indicators (e.g.,
labor force participation, entrepreneurial success, wage and self-employment earnings, business
investment, access to lending); (3) business knowledge and business/management practices; (4)
self-reported and task-based measures of soft skills (e.g., negotiation, persuasion, patience, self-
efficacy, stress, grit, and Big Five).
Three and a half years after the intervention, 88.4% of the original sample was successfully
tracked. However, the attrition rates were slightly higher for the control group than for the two
treatment groups. Recontact rates were 85.3%, 89.7% and 89.3% for the control, hard-skills and
soft-skills groups, respectively (see Table A.3, column (1)). Female and marginally older youth tend
to attrit at higher rates, while youth planning on attending university tend to attrit less. However,
covariates do not appear to explain attrition rates across assignment groups (see Table A.3, column
(2)), with the coefficients on the two treatment dummies being virtually unchanged (nor are their
standard errors) from column (1) to column (2). In other words, differential attrition across groups
is not explained by our rich set of observables, including proxies for abilities, personality traits, and
socio-emotional characteristics.
In order to assess whether differential attrition leads to bias, we retest balance across all of
our 30 outcomes for the estimation sample used in our ITT analysis (Table A.4): only 9 out of
90 comparisons are significantly different from zero; the p-values for joint tests of whether the
30 covariates predict treatment statuses relative to the control are larger than 0.37, consistent
with the notion that the samples remain well-balanced on various youth characteristics 3.5 years
post-intervention.
Finally, Tables A.5 and A.6 document imperfect compliance with assignment among the two
treatment groups, but no contamination of the control group (no defiers). 69% of those assigned to
treatment (SEED hard: 67%; SEED soft: 69%) participated in SEED. Conditional on taking up
treatment, the majority (over 70%) attended 28 or more of the 30 sessions or more. These take-up
8
Topics covered by the hard skills test include general business knowledge (e.g., meaning of an enterprise, defini-
tion of legal ownership, challenges and advantages of self-employment); concepts related to industry sectors, market
competition and its implications; best business practices surrounding sales, marketing, employee management; ele-
ments of profit/loss statements, utility of records keeping, classification of costs; financing strategies to support capital
and business investments; elements of successful business plans; and strategies to identify and develop good business
ideas. The soft skills assessment instead focused on qualities and traits of successful entrepreneurs; communication,
the importance of listening, elements of successful public speaking; negotiation strategies and desirable outcomes; at-
tributes of win-win judgement; personal power (self-mastery, competence, vision, confidence); decision-making skills;
and the importance of teams and social networking.
10
rates are consistent with other entrepreneurship training programs analyzed in the literature, such
as Groh et al. (2015), de Mel et al. (2013) and Calderon et al. (2020), and exceed the 21% take-up
rate in Bandiera et al. (2020).
4.1 Estimation
Identification of program impacts relies on random assignment of applicants to SEED curricula and
control groups. For parsimony and for cleaner interpretation of the results, we present intent-to-
treat (ITT) estimates, which capture the impact of offering SEED training irrespective of actual
take-up.9 Our main estimating equation for the ITTs is given by
yi = α + βh Tih + βs Tis + γXi + i (1)
where yi denotes outcome y for youth i; Tih and Tis correspond to hard skills and soft skills
treatment indicators, respectively; βh and βs represent the ITT for the respective treatment arms;
Xi is a vector of controls. In the paper, we report estimates for which the controls are chosen
using Double/Debiased Machine Learning, DML (Chernozhukov et al., 2018) to maximize preci-
sion.10 DML ITTs and standard errors are estimated using an interactive model in which treatment
dummies are not additively separable.11
When considering multiple hypothesis tests simultaneously, standard statistical techniques will
lead to over-rejection of null hypotheses unless the multiplicity of the testing framework is explicitly
considered. We adopt Romano and Wolf (2005)’s correction, which controls the familywise error
rate (FWER) by taking into account the dependence structure of the test statistics. For a given
family of k conceptually similar blocks of outcomes, the algorithm uses resampling and step-down
methods to estimate the dependence structure of the test statistics and provides corrected p-values.
5 Results
This section organizes outcomes into categories related to their causal distance from the training:
the theory of change posits that the program impacts youth’s skill sets and that these skills, in
turn, shape labor market outcomes and, ultimately, youth’s livelihoods. We first assess the effect of
the training on knowledge of hard and soft skills immediately after the training and 3.5 years after
the intervention. During our follow-up, in addition to hard skills and business knowledge, we also
9
Local Average Treatment Effect, LATE, estimates of the effect of treatment on the treated using initial random
assignment as an instrument are proportional to ITT estimates under a monotonicity assumption (no defiers). That
is, LAT E = IT T /πc , where πc is the fraction of compliers.
10
The main results and overall conclusions are preserved when we omit baseline controls, Xi ; however, as expected,
DML improves precision.
11
DML estimates rely on K-fold cross-fitting (in our case K = 5) to minimize overfitting. In finite samples, the
dependence of the estimator on the particular split creates an additional source of variation. Following Chernozhukov
et al. (2018), to incorporate a measure of this additional source of variation into estimates and their standard errors,
the estimation step is repeated S times (S = 100). Sample medians of the estimates and median standard errors
obtained across the S replications/splits are reported.
11
Table 1: Post-training Knowledge Test Scores
(1) (2) (3) (4) (5) (6) (7) (8)

Hard Skills Assessment (sd) Soft Skills Assessment (sd)
Soft Skills Treatment

β -0.40*** -0.43*** -0.42*** -0.41*** 0.28*** 0.25*** 0.26*** 0.28***
se (0.041) (0.035) (0.033) [0.024] (0.035) (0.034) (0.032) [0.023]
Mean of Hard Skills 0.761 0.465
Observations 1959 1959 1959 1959 1959 1959 1959 1959

Lagged score No Yes Yes Yes No Yes Yes Yes
Student characteristics No No Yes Yes No No Yes Yes
DML No No No Yes No No No Yes
Note: Test scores for the hard and soft skills assessments post intervention are standardized with respect to
knowledge assessment test scores performed prior to training. The mean of the hard skills group corresponds
to the average test score for youth in the hard skill curriculum. All specifications are OLS regressions, with the
exception of columns (4) and (8) where controls are chosen using DML (Chernozhukov et al., 2018).
measure different socio-emotional skills/traits (e.g., Big Five, Stress, Self-Efficacy, etc.) including
task-based measures to assess persuasion and negotiation skills. Finally, we present an overview of
our economic outcomes: labor force participation and business-related outcomes (e.g., enterprise
start-up and survival rate, employment creation, as well as business investment). We conclude by
reporting results on earnings to study to which extent upgrades in entrepreneurship mindset and
skills are rewarded in the labor market and translate into more profitable enterprises.
5.1 Knowledge Assessments, Post Treatment

Students who attended the training completed tests of hard and soft skills knowledge at the begin-
ning and at the the end of the mini-MBA. The pre- and post-training knowledge assessments are
identical, except that the order of questions differs between the tests.
Figure 3 illustrates the pre-to-post shift in the test score distributions for hard and soft skills
knowledge, by treatment arm. Aggregate test scores were computed as standardized weighted
averages following Anderson (2008).12 After three weeks of training, both curricula were associated
with significant improvements in test scores for hard and soft skills knowledge relative pre-program
knowledge.
More importantly, the two curricula led to significant differential knowledge accrual in hard and
soft skills knowledge (see Table 1). Focusing on hard skills knowledge, youth in the hard and soft
skill trainings experienced large knowledge gains, 0.76 sd and 0.36 sd respectively. The hard skill
curriculum increased hard skills test scores by 0.4 sd relative to the soft skill group. This learning
gap remains unchanged when we control for pre-baseline test score performances and additional
12
Test scores for each assessment are standardized with respect to the mean and standard deviation of pre-
treatment test performances.
12
Figure 3: Pre- and Post-Training Knowledge Assessments, by Treatment Arm
(a) Hard Skills Knowledge: Test Scores (z-score)
(b) Soft Skills Knowledge: Test Scores (z-score)
Note: Test scores on soft and hard skills assessments for the two treatment groups are standardized with respect to
the mean and standard deviation of pre-treatment test scores of all youth in the treatment arms.
13
Table 2: Hard Skills Knowledge and Soft Skills – 3.5y Follow-up
(1) (2) (3) (4) (5)

Hard skills
Plasticity Stability Self-efficacy Stress
knowledge
Hard skills Treatment

β 0.131*** 0.130*** 0.134*** 0.052 -0.199***
se [0.040] [0.040] [0.039] [0.040] [0.039]
RW p-value 0.010 0.010 0.010 0.267 0.001
Soft skills Treatment

β 0.130*** 0.175*** 0.065* 0.093** -0.176***
se [0.040] [0.040] [0.039] [0.040] [0.040]
RW p-value 0.010 0.010 0.010 0.267 0.001
Mean control group 0.000 0.000 0.000 0.000 0.000

p-value Hard skills = Soft skills 0.986 0.214 0.110 0.212 0.594
Observations 3893 3893 3893 3893 3893
Note: Results are based on DML (K-fold=5, Splits=100); point estimates and standard errors across splits
are calculated with the median method as in Chernozhukov et al. (2018). RW p-values are computed with the
Romano-Wolf step-down correction, with 1,000 bootstrap repetitions. Plasticity and Stability correspond to the
Big Five metatraits: Plasticity combines Emotional Stability, Agreeableness, and Conscientiousness; Stability
combines Extraversion and Openness.
controls.
In terms of soft skills knowledge, the test scores for both treatment groups improved signif-
icantly, with a 0.28 sd score differential between treatment groups, consistent with the soft skills
intensity of the assigned curricula. Soft Skills assessment test scores increased by 0.74 sd and 0.47
sd for those in the soft skill and hard skill groups, respectively.
5.2 Hard and Soft Skills: 3.5 Years Post-Intervention

Three and a half years after the intervention, meaningful differences in hard skills knowledge and
psycho-social skills/traits remain relative to the control group in both treatment groups (see Table
2).
SEED graduates score better (0.13 sd) on the hard skill knowledge assessment, and exhibit
greater ability to maintain stability and avoid disruption in emotional and social domains (Stability)
and to engage flexibly with novelty, in both behavior and cognition (Plasticity).13 Larger effect
sizes for Stability emerge for the hard skill treatment group (0.13 sd for SEED Hard vs and 0.065
13
The Big Five taxonomy of personality traits (Goldberg, 1993; John et al., 1991) was initially conceptualized as
containing orthogonal dimensions. Recent work has revealed the presence of two higher-order personality traits: (1)
Stability, Self-Control, or Alpha is composed of the shared variance of Emotional Stability (Neuroticism reversed),
Conscientiousness, and Agreeableness; (2) Plasticity, Engagement, or Beta is composed of the shared variance of
Extraversion and Openness/Intellect. For a review of these constructs see De Young (2006), Hirsh et al. (2009) and
citations therein. In Appendix A, Table A.8, we report results for the individual dimensions of Big Five.
14
sd for SEED Soft, both significant at conventional levels). Plasticity coefficients are closer to each
other, with a soft skill estimate of 0.175 sd and a hard skill treatment of 0.13 sd (the two effects not
being statistically distinguishable). The psycho-social skills gain extends to youth’s ability to better
handle stress, with estimated effects of 0.2 sd and 0.18 sd for hard and soft curricula, respectively.
While the gaps in intrapersonal skills (as proxied by Plasticity, Stability, and Stress management
skills) between the two treatments closed 3.5 years post-intervention, some important differences
remain. These favor the soft skills treatment group and emerge for self-efficacy 14 (the belief about
one’s ability to effectively perform tasks, including interpersonal tasks, needed to achieve a valued
goal) and interpersonal soft skills in the context of negotiation and persuasion task-based measures.
5.2.1 Task-Based Measures: Negotiation and Persuasion
When founding, running and growing a venture, entrepreneurs constantly need to negotiate. They
regularly have to settle agreements with various stakeholders to acquire human and financial re-
sources, making negotiation skills central to entrepreneurial success (Artinger et al., 2015). To
assess whether the newly acquired skills can help youth in their entrepreneurial endeavors by pro-
viding them with better tools to negotiate more favorable prices or to persuade and convincingly
present their business plans and point of view, we turn attention to negotiation and persuasion.
The defining characteristic of negotiation is the existence of two parties who share an impor-
tant objective but have conflicting goals or interests (Bazerman and Neale, 1994). Negotiation is
the process of interpersonal communication in which interested parties resolve disputes, agree upon
courses of action, bargain for individual or collective advantage, and/or attempt to craft outcomes
which serve their mutual interests. Effective negotiation involves skills such as communication,
persuasion, planning, strategic thinking, and cooperating (Thompson (1990) and Bazerman et al.
(2000)). To assess study participants’ negotiation skills, we included a lab field experiment with
the potential to generate real payoffs for both players (Bontempo and Iyengar, 2008).
The Buyer plays the role of Commissioner of the Ministry of Lands, who is tasked to purchase
Butagira Farms to construct a rail line. The Seller plays the role of the owner of Butagira Farms.
The Buyer (the enumerator) and Seller (the respondent) have already agreed that the land is worth
UGX 600 million (approximately $573,295.75, PPP adjusted), but the two will need to agree on (1)
financing terms, i.e., how much is paid upfront, with the remainder being paid in equal installments
over a 12-month period (Table 3, Column 2); and (2) the project start date between two and 12
months (Table 3, Column 3).
Both Buyer and Seller are informed that they will be entered into a lottery with a $400 prize,
with a probability of winning that is as a function of the number of points obtained in the final
deal based on the payoff tables in Figure 4.15 The Seller and Buyer have four minutes to reach an
14
Self-efficacy beliefs determine how people feel, think, motivate themselves and behave (Bandura, 2010). Self-
efficacy does not refer to one’s abilities but to how strongly one believes one can use one’s abilities to work toward
goals (Latham and Locke (2007) and Bandura (2010)). Self-efficacy is not a unitary construct or trait.
15
Figure 4 also summarizes the Buyer’s and Seller’s ideal outcomes and strategies. The Buyer prefers a down
payment as small as possible and for the project to start in two months. The Buyer assigns relatively higher weight
15
Figure 4: Negotiation: Payoff Tables
Note: Buyer’s and Seller’s strategies: the Seller, the respondent, prefers a full payment, UGX 600 million, up front
and a delayed project start date; the Seller places relatively more weight on the financing arrangement than the
starting date. The Buyer prefers a small down payment but an early start date; the Buyer assigns higher weight to
the start date than to the financing arrangements.
agreement, which is the only scenario that will allow both of them to enter the lottery.
Table 3 summarizes these key outcomes. Youth in the soft skills-focused SEED exhibit better
negotiation strategies not only relative to the control group but also relative to those in hard
skills-focused SEED. The Seller’s (respondent) best outcome corresponds to full payment up front
(down payments equal to UGX 600 million) and a project start date in 12 months.16 Negotiation
strategies do not appear to influence the likelihood of agreement, which is fairly high even for
the control group (81%); however, youth in the soft-skill SEED strike better/more advantageous
deals for themselves: later starting dates (7%, approximately one week later) and higher down
payments (9%, or approximately $28,892 which is roughly 5% of the value of the land), relative to
the control group. There are no statistically significant impacts for youth who attended the hard
skills treatment.
At the core of negotiation skills is persuasion; that is, the process of changing (without duress)
a person’s opinion , feelings, behaviors, or general evaluations (attitudes) toward some object,
issue, or person (Cialdini, 2001). Communication, emotional intelligence, active listening, logic,
and reasoning are some of the key skills for persuasion (Cialdini, 2018).17
to the starting date (more points) than to the financing arrangement (fewer points). The Seller prefers to a full
payment, UGX 600 million, up front and a project start date in 12 months. The Seller places more weight (more
points) on the financing arrangement than the starting date (fewer points).
16
In a separate paper, we study in detail the negotiation strategies: youth in the soft treatment arm engage in
lengthier conversations and provide longer arguments, (as proxied by the number of words used by both Sellers and
Buyers), and engage in more rounds of offers and counter-offers.
17
Cialdini (2018) describes six principles that are related to persuasion: (1) Liking – people like those who like
them (uncover real similarities and offer genuine praise); (2) Reciprocity – people repay in kind (give what you want
16
Table 3: Task-based Measures: Negotiation and Persuasion
(1) (2) (3) (4) (5)

Negotiation Persuasion
Final agreement Final agreement
Agreement on How Would like
down payment on start date
final offer convincing to hire
(log) (log month)

β -0.028 0.027 -0.027 0.007 0.003
se [0.017] [0.025] [0.025] [0.006] [0.007]
RW p-value 0.399 0.618 0.618 0.618 0.675

β -0.022 0.090*** 0.070*** 0.014** 0.015**
se [0.017] [0.024] [0.025] [0.007] [0.007]
RW p-value 0.190 0.006 0.035 0.099 0.096
Mean control group 0.814 321,017 3.749 0.022 0.026

Observations 3689 2938 2938 3719 3719
Romano-Wolf step-down correction, with 1,000 bootstrap repetitions. Outcomes in columns (1), (4) and (5) are
dichotomous variables.
During the persuasion task, the participant was asked to convince a group of hypothetical
government officials who intend to distribute more land to farmers that you should be granted
more land which, in turn, will make your business more profitable. Youth in the study were
allotted one minute to make a persuasive pitch and their conversations were audio recorded. To
assess the quality of the case they made, MBA students and business professionals recruited in
Uganda were asked to score youth’s performance during the persuasion task.
Youth in the soft skills treatment were more persuasive than youth in the control group. In
particular, they were more likely to be identified as convincing (1.4 percentage point, pp, or equiv-
alent to 64% more convincing), more likely to be someone whom the evaluators would hire (1.5 pp,
or 57% more likely to be hired). Consistent with the results from the negotiation task, there are
no significant impacts for the hard skill treatment group relative to the control group. However,
we are unable to reject the null of equal impacts for the two treatment groups.
Overall, the findings in this section suggest that 3.5 years post-intervention, treatment groups
performed better on hard skills knowledge assessments and demonstrated better soft skills relative
to the control group. Furthermore, differentials in hard skills knowledge narrowed among youth
to receive); (3) Social proof – people follow the lead similar to that of others (use peer power whenever it’s available);
(4) Consistency – people align with their clear commitments (make their commitments active, public, and voluntary);
(5) Authority – people defer to experts (expose your expertise; don’t assume it’s self-evident); and (6) Scarcity –
people want more of what they can have less of (highlight unique benefits and exclusive information).
17
attending the two curricula, as did the gap in intrapersonal skills such as Stability, Plasticity,
and Stress. However, some skill differentials have persisted: youth in the soft skill treatments
demonstrated higher self-efficacy and more mastery of skills regulating interpersonal behavior and
performance, such as negotiation and persuasion.
5.3 Economic Outcomes

We now turn to economic outcomes to better understand whether the newly acquired skills and
business knowledge translated into business success and/or how they are rewarded in the labor
market.
Youth in our sample are in their mid-20s and a large share (around 70%) are economically
active (that is, either working for someone else or self-employed). Yet, SEED graduates are active
at higher rates, with hard and soft skill treatment groups participating 6.2 pp (9.8%) more and
4.4 pp (7.1%) than the control group, respectively (see Table 4). Roughly 43% of the control
group is currently working for someone else (i.e., dependent work). Youth in SEED hard and
SEED soft are largely engaged in dependent work at similar rates, with those in SEED hard being
engaged at marginally higher rates (3.8 pp, significant at the 10% level). As such, the difference
in rates of economic activity is almost entirely driven by higher degrees of entrepreneurship among
program participants. Hard and soft treatment groups are 19% (approximately 6 pp, for each
treatment group) more likely to be self-employed. SEED graduates are also more likely to be
contemporaneously self-employed and working for someone-else.
Table 4: Labor Market Outcomes: Employment Status
(1) (2) (3) (4)

Economically Currently working Currently Currently working
Active for someone else self employed and self employed

β 0.066*** 0.038* 0.063*** 0.035***
se [0.018] [0.020] [0.019] [0.013]
RW p-value 0.002 0.074 0.008 0.018

β 0.048*** 0.009 0.066*** 0.029**
se [0.019] [0.020] [0.019] [0.013]
RW p-value 0.045 0.662 0.006 0.050
Mean control group 0.672 0.431 0.336 0.095

Observations 3891 3891 3890 3891
p-value Hard skills = Soft skills 0.490 0.320 0.907 0.746
Romano-Wolf step-down correction, with 1,000 bootstrap repetitions. All outcome variables are dichotomous;
economically active is defined either working for someone else or self employed.
18
Table 5: Business creation and investment
(1) (2) (3) (4) (5)

Number of Business
Ever started Business Number of
businesses investment
a business currently open employees (log)
started (log)

β 0.105*** 0.105*** 0.063*** 0.121*** 0.429***
se [0.020] [0.019] [0.019] [0.034] [0.119]
RW p-value 0.001 0.001 0.008 0.008 0.004

β 0.112*** 0.115*** 0.066*** 0.144*** 0.360***
se [0.020] [0.019] [0.019] [0.034] [0.119]
RW p-value 0.001 0.001 0.002 0.001 0.008
Mean control group 0.474 0.725 0.336 1.810 1276

Observations 3890 3890 3890 3893 3890
Romano-Wolf step-down correction, with 1,000 bootstrap repetitions. Outcomes in columns (4) and (5) are not
conditional on employment status; number of employees (business investment) is defined as the joint outcome of
owning a business and number of employees (business investment).
Indeed, SEED graduates are not only more likely to start a business (10.5 and 11.1 pp, respec-
tively) relative to youth in the control group, 47.4% of whom started a business, but they also
started more businesses and are more successful in ensuring their survival (see 5). 33.6% of the
control group owned one or more business currently open at the time of our 3.5 year data collection.
Youth in the treatment groups are 18% (approximately 6 pp) more likely to own currently open
businesses, with no statistically significant difference between treatment groups.
The entrepreneurial efforts of SEED graduates also led to job creation (as proxied by the number
of employees, hired by currently open enterprises) and yielded businesses with higher potential for
growth and expansion (as proxied by business investment/capital). The last two columns of Table
5 capture these dimensions.
Businesses (open at the time of the 3.5 year follow-up) owned by youth enrolled in SEED gener-
ated 12.1% (SEED hard) and 14.4% (SEED soft) more jobs relative to the control group, as proxied
by dependent employees. To put these figures into perspective, currently open businesses owned
by SEED graduates were responsible for 63% of total employees in our sample; the program led
to roughly 379 additional (current) jobs. These job figures exclude business owners and employees
who were hired by businesses that were not open during the medium run follow-up. Furthermore,
SEED graduates seeded enterprises that attracted considerably more capital investments than the
control group: 42.9% and 36% for the hard and soft treatment groups, suggesting that SEED-led
19
Figure 5: Cumulative Distribution Functions, CDFs, for Selected Variables
(a) log Total Current Earnings (b) log Wage Earnings
Hard skills vs Control, KS p-value = 0.000; Hard skills vs Control: KS p-value = 0.580;
Soft skills vs Control, KS p-value = 0.000; Soft skills vs Control: KS p-value = 0.840;
Hard skills vs Soft skills, KS p-value = 0.240. Hard skills vs Soft skills: KS p-value = 0.680.
(c) log Profits (d) log Business Investments (Capital)
Hard skills vs Control KS p-value = 0.080; Hard skills vs Control, KS p-value = 0.000;
Soft skills vs Control, KS p-value = 0.070; Soft skills vs Control, KS p-value = 0.010;
Hard skills vs Soft skills, KS p-value = 0.990. Hard skills vs Soft skills, KS p-value = 0.920.
Note: All variables are expressed in log of USD PPP adjusted. Below each graph we report p-values (KS p-value)
for the corresponding pairwise CDFs Kolmogorov-Smirnov test.The KS test is invariant to logarithmic and other
transformations of the data
20
ventures may be of higher quality/potential.18 Figure 5d compares the CDFs of log business in-
vestments (capital): the distributions of the two treatments groups are indistinguishable, but are
statistically different from that of the control group. The gap results largely from a differential mass
at zero business investment, either because youth in the control group have no capital investment
conditional on owning a business or because they do not own an enterprise. The hard and soft
treatment groups are 27% and 23% more likely to report positive capital investments relative to
the control group.
5.3.1 Total Earnings, Wage Earnings, and Profits
Taken altogether, SEED graduates possess better knowledge and better intra- and inter-personal
skills to succeed in the labor market; overall, they are more likely to start businesses that survive
over time and higher quality ventures that create employment opportunities and attract larger
capital investments. A complementary set of outcomes pertains to the extent to which the upgrades
in entrepreneurship mindset and skills are rewarded in the labor market and translate into more
profitable enterprises (see Table 6).
Monthly (current) earnings from all sources for both treatment groups significantly outpace
those of the control group; namely, earnings for the hard and soft skill treatments are 32.1% and
29.8% higher, respectively. A closer examination of the total earnings CDF (Figure 5a) suggests
that differences in earnings begin to emerge around the median, when earnings in the control group
start lagging behind those of the treatment groups.
Segregating between wage earnings and business earnings (profits),19 , the aforementioned earn-
ings gap between treatment and control groups is almost entirely explained by business profits:
SEED graduates report 24.2% (hard) and 27.2% (soft) higher profits than their counterparts in the
control group.
Previous research has noted that business skills training had limited impact on businesses’
performance. However, interventions that are tailored to participants’ needs or foster an en-
trepreneurial mindset have shown more promise. The impacts on profits reported here are compa-
rable to those of Campos et al. (2017), where training increased profits by 30%, the only other study
focusing on soft-skills/personal initiative training that detected positive and significant impacts on
profits (McKenzie, 2020).
Taken altogether, our results point to the fact that SEED training was successful in jump-
starting youth human capital (soft and hard skills) accumulation to set them on a higher earning
trajectory and expanded income opportunities. Moreover, the program contributed to the creation
of vibrant ventures, with SEED graduates behaving like transformational/opportunity entrepreneurs
18
Youth in the soft skills-focused training were also more likely to start a formal business (3 pp, 23%) relative to
the control group.
19
Monthly profits (wage earnings) are defined as the joint outcome of owning a business (working for someone
else) and current monthly profits (earnings). ITT estimates for outcomes that are conditional on employment status
(e.g., defined only for entrepreneurs and/or wage workers) would be biased and inconsistent, since it would imply
conditioning on an outcome that is endogenous to treatment.
21
Table 6: Earnings
(1) (2) (3)

Current Business Wage
Total Earnings Profits Earnings

β 0.321** 0.242** 0.094*
se [0.089] [0.101] [0.053]
RW p-value 0.005 0.058 0.092

β 0.298*** 0.272*** 0.034
se [0.091] [0.101] [0.053]
RW p-value 0.009 0.028 0.558
Mean control group (USD PPP adjusted) 229.58 195.83 6.25

p-value Hard skills = Soft skills 0.854 0.831 0.427
Observations 3890 3890 3890
Romano-Wolf step-down correction, with 1,000 bootstrap repetitions. Outcomes in columns (2) and (3) are not
conditional on employment status; monthly business profits (monthly wage earnings) are defined as the joint
outcome of owning a business (wage work) and profits (wage earnings).
(as in Schoar, 2010; Ardagna and Lusardi, 2010): that is, they were more successful in creating
dynamic businesses that grown beyond the scope of an individual’s subsistence needs and provide
jobs and income for others.
5.4 Costs and Benefits

Cost analysis falls into two broad categories: Cost-Benefit Analysis and Cost Effectiveness. CE
captures the amount of “effect” a program achieves on one outcome for a given cost, while cost-
benefit analysis combines all the different costs and benefits of a program into a single scale (usually
a monetary metric) and shows the ratio of combined benefits to combined costs (Dhaliwal et al.,
2012).
The conservative measure of benefits that we consider in this section is youths’ monthly earnings
3.5 years post-intervention. This is a conservative measure of benefit since it abstracts from any
benefits that are realized since the intervention as well as from any impacts on mental well-being
or those resulting from job creation and social spillovers, all of which are difficult to monetize.
This said, the non-monetary benefits of such skills training programs – and of SEED training, in
particular – can be quite sizable.20 A medium run evaluation of another secondary school social
20
SEED graduates report being more satisfied with their quality of life (0.16 sd and 0.17 sd for hard- and soft-skills
treatments, respectively); as documented earlier, they also report less stress.
22
leadership intervention carried out by Educate!, our implementing partner for SEED, suggests even
larger non-monetary impacts.21
The research and evaluation team was directly involved in the implementation and supervision
of the SEED program, as such we have access to detailed information on intervention costs. Total
intervention costs were estimated at $118 per student.22 To put these numbers into perspective,
based on McKenzie (2020)’s random effects meta-analysis, traditional (hard skills only) training
programs yield positive average effects on both profits (10.1%) and sales (4.6%). Costs for these
program range from $3,537, or US$177 per participant. For a firm earning US$100 a month, a
10% increase in profits would recover the $177 cost of a training within 18 months McKenzie et al.
(2020).
Campos et al. (2017)’s personal initiative training, which is more directly comparable to SEED’s
curricula, it cost US$756 per invited participant and yielded a $60-per-month increase in monthly
profits over the first two years. Youth in the hard and soft skill trainings earned an additional
$73.47 and $68.41 per month, respectively. Therefore, if SEED’s program benefits are proxied by
treatment effects on monthly earnings 3.5 years post-intervention, two months of earnings alone
would exceed the cost of the program. These benefits abstract from the job- and business-creation
benefits of the program, which were substantial: relative to the control group, SEED entrepreneurs
created 985 additional (employee) jobs and 550 new businesses.
6 Concluding Remarks
In this paper, we study the medium-term impacts of SEED, which was implemented at-scale in
Uganda. SEED is an innovative in-residence 3-week mini-MBA modeled after western business
school curricula. The program featured two separate treatments, with the hard-skills MBA featur-
ing a mix of approximately 75% hard skills and 25% soft skills and the soft skills curriculum having
the reverse mix.
SEED training was successful not only at jump-starting youth human capital accumulation
(both soft and hard skills), but also at setting them on higher earnings trajectories and expanding
their income opportunities. In particular, SEED was a catalyst for the inception of dynamic and
more profitable ventures which attracted larger capital investments, and expanded employment
opportunities, creating over 985 additional (employee) jobs. Our results point to SEED graduates
behaving like transformational/opportunity entrepreneurs (as in Schoar, 2010). The intervention
was highly cost-effective and compares favorably to existing training alternatives that tend to be
more expensive and often yield lower benefits.
Two critical program features support SEED’s effectiveness. First, SEED was designed and
21
The Educate! program generated substantial social spillovers along several dimensions: delayed family formation,
lower incidence of risky behaviors, shifts in gender-related social norms, and reductions in intimate partner violence.
22
Total costs included salaries and benefits for trained teachers (10%), curriculum development and testing (2%),
teacher recruitment and training (4%), site management staff costs (4%), venue, transports, meal, staff plus other stu-
dents costs (67%), training materials and supplies (5%), supervision and support to teachers (2%), and administration
country office (5%).
23
implemented as a high quality intervention. The curricula were informed by rigorous evidence from
the economic and psychology literatures, which have been able to identify which skills/personality
traits matter in shaping high performing entrepreneurs. Furthermore, teachers received extensive,
high quality training on teaching methods and soft and hard skills knowledge. Implicitly, our
results corroborate the proposition that high quality teacher and youth trainings can be delivered
at relatively low costs.
Second, SEED targeted youth at the end of secondary school, which is earlier in life than tra-
ditional training interventions. This critical age and developmental stage has garnered increasing
attention. The period of adolescence through to mid-20s represents a critical stage of brain devel-
opment, featuring high degrees of plasticity in those areas of the brain that are responsible for the
regulation of automatic or instinctive reactions, risk-taking behavior, self-control, reflective reason-
ing, communication, among others. In fact, research increasingly points to youth as the key phase
during which those skills that are at the core of transformational entrepreneurship can be shaped.
It is also worth highlighting that youth in our study are positively selected: not only were they
in their last year of secondary school, but the majority also planned to attend university, both of
which correlate with higher ability. Larger effects for higher potential individuals is consistent with
the calls for better targeting emerging from recent work. Having estimated no program effects for
business training or workfare programs, some authors examined impact heterogeneity and concluded
that a more careful targeting of businesses could improve the impacts these programs (e.g., Bardasi
et al., 2021; Bertrand et al., 2021, and citations therein). The efficacy of SEED for youth with
lower levels of education/ability remains untested and is an avenue we are planning on exploring.
Several important questions remain, which we hope to answer with our 8-year follow-up. As
McKenzie and Woodruff (2017) points out, most of the existing literature on business training does
not examine the effects of the training at different points in time. The few exceptions indicate that
some effects disappear in the longer term (e.g., Blattman, Fiala and Martinez, 2020; Blattman,
Dercon and Franklin, 2019). Furthermore, while SEED explicitly features soft skills (with two
distinct intensities) as a centerpiece of its curriculum, we only observe differences in inter-personal
skills and self-efficacy between the two treatment arms, but no statistical differences overall for
economic outcomes. Will this pattern persist in long-term or are returns to inter-personal skills
just slower to emerge?
SEED-like interventions call for future research to gain insight into their optimal timing (how
late or how early can these skills be taught?), into their complementarities with other initiatives
(e.g., start-up capital, mentoring) and into potential social spillovers of these skills on non-economic
outcomes (e.g., gender empowerment). Between 2010 and 2030, global labor markets will need to
generate roughly three-quarters of a billion new jobs (Bloom et al., 2018), an especially daunt-
ing task in light of growing young populations and often-underperforming education systems –
themselves under additional stress during the pandemic. As a result, training programs designed
to ”make” transformational entrepreneurs capable of generating positive ripple effects for the econ-
omy are critical investments to face this development challenge.
24
A Appendix: Tables
Table A.1: Baseline Balance Table: Baseline Sample
Control Hard Skills Soft Skills

p-values
(n=1,198) (n=1,613) (n=1,591)
Hard vs Soft vs Hard vs

Mean Mean Mean
Control Control Soft
Demographics and Expectations
Age at baseline 20.04 20 20 0.423 0.429 0.994
Female 0.34 0.36 0.36 0.204 0.219 0.968
Boarding student 0.74 0.73 0.73 0.69 0.619 0.914
Business work experience index (Anderson) 0.02 -0.01 -0.01 0.107 0.184 0.761
Future subjective personal wealth (z-score) -0.02 0.02 0.00 0.304 0.524 0.676
Plans to attend university (indicator) 0.71 0.7 0.71 0.498 0.885 0.566
Family background
Parental SES index (Anderson) 0.00 -0.01 0.00 0.675 0.994 0.646
Housing index (Anderson) -0.01 0.01 0.00 0.51 0.683 0.787
Subjective family weath (z-score) -0.02 0.00 0.01 0.592 0.379 0.709
Big 5 (z-score)
Extraversion 2.73 2.75 2.7 0.545 0.405 0.12
Openness 4.15 4.13 4.13 0.494 0.361 0.801
Emotional stability 3.86 3.88 3.85 0.473 0.891 0.356
Agreeable 3.63 3.63 3.61 0.908 0.588 0.477
Conscientious 3.88 3.9 3.88 0.612 0.892 0.488
Skills (z-score)
Memory (# digits recited backwards) 0.04 -0.01 -0.02 0.152 0.138 0.953
Intelligence (# Raven’s matrices correct) -0.01 0.02 -0.01 0.326 0.841 0.399
UACE O-levels score -0.04 0.03 0.00 0.061 0.223 0.479
Math and business knowledge index 0.01 0.01 -0.02 0.99 0.339 0.307
Leadership index -0.03 0.00 0.04 0.345 0.071 0.349

Perceived control index 0.00 0.01 0.01 0.938 0.789 0.836
Self-confidence index 0.05 -0.04 0.01 0.016 0.249 0.176
Prosocial index -0.05 0.00 0.05 0.169 0.013 0.226
Patience index -0.02 0.02 0.01 0.277 0.385 0.815
Willingness to take risks 0.00 0.00 0.00 0.959 0.914 0.864
Teacher and peer acceptance index -0.03 0.01 0.03 0.248 0.111 0.633
Anxiety and depression index -0.05 0.00 0.04 0.166 0.017 0.279
Travel time (hours)

School to hard skills host school 1.49 1.51 1.50 0.583 0.791 0.759
School to hard skills meeting point 1.31 1.33 1.32 0.502 0.667 0.795
school to soft skills host school 1.21 1.24 1.23 0.434 0.603 0.779
School to soft skills meeting point 1.14 1.17 1.16 0.503 0.661 0.804
F-test of joint significance (F-stat) 1.27 1.20 0.59

p-value 0.167 0.221 0.951
25
Table A.2: Baseline Balance Table: Take-up Sample and Control Group
Control Hard skills Soft skills

p-values
(n=1,198) (n=1,089) (n=1,105)
Hard vs Soft vs Hard vs

Mean Mean Mean
Control Control Soft
Age at baseline 20.04 20.07 20.06 0.692 0.77 0.917
Female 0.34 0.32 0.33 0.446 0.588 0.827
Boarding student 0.74 0.70 0.71 0.035 0.139 0.532
Business work experience index (Anderson) 0.02 -0.01 0.00 0.159 0.358 0.629
Future subjective personal wealth (z-score) -0.02 -0.01 0.00 0.669 0.561 0.882
Family background
Parental SES index (Anderson) 0.00 -0.02 -0.02 0.352 0.336 0.978
Housing index (Anderson) -0.01 -0.06 -0.05 0.158 0.26 0.776
Subjective family weath (z-score) -0.02 -0.04 -0.04 0.636 0.685 0.947
Big 5 (z-score)
Extraversion 2.73 2.76 2.72 0.382 0.822 0.282
Openness 4.15 4.11 4.14 0.228 0.751 0.383
Agreeable 3.63 3.6 3.59 0.303 0.245 0.901
Conscientious 3.88 3.87 3.88 0.634 0.771 0.855
Skills (z-score)
Memory (# digits recited backwards) 0.04 -0.06 -0.06 0.014 0.017 0.934
UACE O-levels score -0.04 0.05 0.02 0.03 0.129 0.516
Math and business knowledge index 0.01 0.01 -0.02 0.887 0.354 0.444

Prosocial index -0.05 0.00 0.05 0.296 0.022 0.225
Patience index -0.02 0.04 0.02 0.16 0.339 0.657
Willingness to take risks 0.00 0.00 0.02 0.92 0.616 0.556
Teacher and peer acceptance index -0.03 0.02 0.02 0.164 0.205 0.901
Anxiety and depression index -1.17 -1.19 -1.18 0.041 0.39 0.244

p-value 0.013 0.247 0.200
26
Table A.3: Attrition
(1) (2)
β se β se
Hard Skills Treatment -0.044*** (0.012) -0.044*** (0.012)
Soft Skills Treatment -0.040*** (0.012) -0.038*** (0.012)

Age at baseline 0.009** (0.004)
Female 0.025** (0.011)
Boarding student -0.016 (0.012)
Business work experience index (Anderson) -0.005 (0.011)
Future subjective wealth (z-score) -0.010** (0.005)
Plans to attend university (indicator) -0.028** (0.011)
Family background
Parental SES index (Anderson) 0.020** (0.010)
Housing index (Anderson) -0.021*** (0.007)
Subjective family weath (z-score) 0.000 (0.005)
Big 5 (z-score)
Extraversion -0.010* (0.006)
Openness -0.01 (0.007)
Emotional stability 0.005 (0.007)
Agreeable 0.008 (0.008)
Conscientious 0.006 (0.008)
Skills (z-score)
Memory (# digits recited backwards) -0.012** (0.005)
Intelligence (# Raven’s matrices correct) -0.002 (0.005)
UACE O-levels score 0.003 (0.005)
Math and business knowledge index -0.007 (0.005)
Leadership index -0.010* (0.006)

Perceived control index 0.000 (0.005)
Self-confidence index -0.002 (0.006)
Prosocial index -0.004 (0.006)
Patience index 0.005 (0.005)
Willingness to take risks 0.002 (0.005)
Teacher and peer acceptance index -0.003 (0.006)
Anxiety and depression index -0.007 (0.006)
Travel time
School to hard skills host school 0.024* (0.013)
School to hard skills meeting point 0.009 (0.015)
school to soft skills host school -0.003 (0.020)
School to soft skills meeting point -0.009 (0.024)
Observations 4,402 4,402
27
Table A.4: Baseline Balance Table: Estimation Sample, 3.5 year Follow-up
Control Hard skills Soft skills

p-values
(n=1,022) (n=1,447) (n=1,421)
Soft vs Soft vs Hard vs

Mean Mean Mean
Hard Control Control
Age at baseline 20.00 19.98 20.00 0.735 0.981 0.692
Female 0.32 0.36 0.35 0.036 0.094 0.647
Boarding student (indicator) 0.75 0.73 0.74 0.390 0.483 0.865
Business work experience index (Anderson) 0.02 -0.01 0.00 0.229 0.311 0.839
Future subjective wealth (z-score) 0.01 0.03 0.01 0.548 0.998 0.509
Family background
Parental SES index (Anderson) 0.00 -0.01 0.00 0.730 0.749 0.466
Housing index (Anderson) 0.01 0.02 0.01 0.834 0.929 0.896
Subjective family wealth (z-score) 0.00 0.00 0.02 0.966 0.629 0.629
Big 5 (z-score)
Extraversion 2.74 2.76 2.71 0.750 0.358 0.174
Openness 4.15 4.15 4.14 0.843 0.547 0.657
Agreeable 3.62 3.62 3.62 0.997 0.964 0.964
Conscientious 3.89 3.90 3.88 0.686 0.809 0.478
Skills (z-score)
Memory (# digits recited backwards) 0.05 0.00 -0.02 0.183 0.085 0.663
UACE O-levels score -0.07 0.02 -0.01 0.037 0.149 0.484
Math and business knowledge index 0.01 0.02 0.00 0.804 0.677 0.466

Prosocial index -0.03 0.02 0.05 0.285 0.046 0.308
Patience index -0.03 0.01 0.01 0.282 0.359 0.866
Willingness to take risks -0.01 0.01 0.01 0.510 0.584 0.905
Teacher and peer acceptance index 0.00 0.02 0.03 0.639 0.449 0.750
Anxiety and depression index -0.05 0.00 0.04 0.232 0.046 0.378
Travel time (hours)

School to hard skills host school 1.43 1.47 1.46 0.304 0.440 0.782
School to hard skills meeting point 1.25 1.30 1.30 0.222 0.263 0.915
School to soft skills host school 1.18 1.22 1.24 0.354 0.192 0.677
School to soft skills meeting point 1.11 1.15 1.17 0.359 0.185 0.650

p-value 0.368 0.444 0.995
28
Table A.5: Take-up and Completion Rates
(1) (2) (4) (5)
Any Attendance Full Attendance

Hard skills Soft skills Hard skills Soft skills
β se β se β se β se

Age at baseline 0.003 (0.009) 0.003 (0.009) 0.002 (0.010) 0.003 (0.010)
Female -0.083*** (0.028) -0.075*** (0.028) -0.069** (0.030) -0.061** (0.030)
Boarding student -0.107*** (0.028) -0.045 (0.028) -0.054* (0.030) -0.060* (0.031)
Business work experience index (Anderson) -0.018 (0.028) 0.012 (0.029) 0.004 (0.030) -0.001 (0.031)
Future subjective personal wealth (z-score) -0.002 (0.013) 0.015 (0.013) 0.002 (0.013) 0.009 (0.014)
Plans to attend university (indicator) -0.093*** (0.027) 0.014 (0.028) -0.054* (0.029) 0.014 (0.030)
Family background
Parental SES index (Anderson) 0.038 (0.025) -0.012 (0.025) -0.001 (0.027) 0.013 (0.027)
Housing index (Anderson) -0.061*** (0.016) -0.027* (0.016) -0.036** (0.017) -0.014 (0.018)
Subjective family weath (z-score) -0.002 (0.013) -0.027** (0.013) 0.016 (0.014) -0.014 (0.014)
Big 5 (z-score)
Extraversion 0.013 (0.013) 0.013 (0.014) -0.014 (0.014) -0.001 (0.015)
Openness -0.025 (0.018) 0.023 (0.018) -0.001 (0.019) 0.026 (0.019)
Emotional stability -0.001 (0.018) 0.008 (0.018) -0.004 (0.020) 0.009 (0.019)
Agreeable -0.033* (0.020) -0.012 (0.019) -0.035* (0.021) -0.002 (0.021)
Conscientious -0.039** (0.018) -0.010 (0.019) -0.037* (0.020) -0.013 (0.021)
Skills (z-score)
Memory (# digits recited backwards) -0.039*** (0.012) -0.035*** (0.012) -0.028** (0.013) -0.040*** (0.013)
Intelligence (# Raven’s matrices correct) 0.015 (0.013) -0.004 (0.013) 0.027* (0.014) -0.008 (0.014)
UACE O-levels score 0.007 (0.013) 0.009 (0.013) -0.006 (0.014) 0.021 (0.014)
Math and business knowledge index 0.003 (0.012) -0.003 (0.012) 0.01 (0.013) 0.006 (0.013)
Leadership index 0.026* (0.014) -0.013 (0.014) 0.028* (0.015) -0.009 (0.015)
Perceived control index -0.002 (0.013) -0.006 (0.013) -0.011 (0.014) 0.001 (0.014)
Self-confidence index 0.016 (0.013) 0.01 (0.014) -0.008 (0.014) -0.006 (0.015)
Prosocial index 0.001 (0.014) 0.009 (0.015) -0.003 (0.015) 0 (0.016)
Patience index 0.001 (0.013) 0.001 (0.013) -0.005 (0.014) -0.008 (0.014)
Willingness to take risks 0.007 (0.012) 0.008 (0.013) -0.009 (0.013) -0.011 (0.014)
Teacher and peer acceptance index 0.01 (0.014) -0.011 (0.014) -0.009 (0.016) -0.021 (0.015)
Anxiety and depression index 0.018 (0.014) -0.01 -0.013 0.003 (0.015) 0.003 (0.014)
Constant 1.126*** (0.237) 0.576** (0.240) 0.820*** (0.255) 0.343 (0.259)
Treatment group take-up rate 0.675 0.695 0.442 0.449

Observations 1,612 1,591 1,612 1,591
R-squared 0.095 0.068 0.069 0.067
29
Table A.6: Compliers and Defiers
No attendance Any Attendance Total

Control group 1,022 0 1,022
T1: Hard skills 445 1,002 1,447
T2: Soft skills 409 1,012 1,421
Total 1,876 2,014 3,890
Table A.7: Knowledge Hard Skills: Disaggregated Elements)
(1) (2) (3) (4) (5) (6)

Opportunities
Effects of Cost Utility of Profit/loss Hard skills
for generating
competition categorization record keeping statements index
business ideas

β 0.116*** 0.062 0.127*** 0.033 -0.007 0.131***
se [0.041] [0.040] [0.041] [0.041] [0.041] [0.040]
RW p-value 0.027 0.443 0.013 0.695 0.730 -

β 0.132*** 0.131*** 0.068* -0.010 0.006 0.130***
se [0.041] [0.041] [0.041] [0.041] [0.041] [0.040]
RW p-value 0.011 0.011 0.303 0.982 0.990 -
Mean control group 0.000 0.000 0.000 0.000 0.000 0.000

p-value Hard skills = Soft skills 0.779 0.237 0.301 0.460 0.832 -
Observations 3893 3893 3893 3893 3893 3893
Romano-Wolf step-down correction, with 1,000 bootstrap repetitions.
30
Table A.8: Big Five, Individual Dimensions
(1) (2) (3) (4) (5)

Emotional
Agreeable Conscientious Extrovert Openness
Stability

β 0.088** 0.115*** 0.130*** 0.115*** 0.078*
se [0.039] [0.040] [0.040] [0.040] [0.040]
RW p-value 0.065 0.026 0.025 0.026 0.065

β 0.062 0.055 0.157*** 0.038 0.122***
se [0.039] [0.040] [0.040] [0.040] [0.040]
RW p-value 0.242 0.270 0.002 0.370 0.008
Mean control group 0.000 0.000 0.000 0.000 0.000

p-val Hard skills = Soft skills 0.634 0.282 0.627 0.178 0.435
Observations 3893 3893 3893 3893 3893
Romano-Wolf step-down correction, with 1,000 bootstrap repetitions.
31
References
Ahmed, S. P., Bittencourt-Hewitt, A. and Sebastian, C. L. (2015), ‘Neurocognitive bases of emotion
regulation development in adolescence’, Developmental Cognitive Neuroscience 15, 11–25.
URL: https://www.sciencedirect.com/science/article/pii/S1878929315000717
Alan, S., Boneva, T. and Ertac, S. (2019), ‘Ever Failed, Try Again, Succeed Better: Results from
a Randomized Educational Intervention on Grit*’, The Quarterly Journal of Economics .
Alan, S. and Ertac, S. (2018), ‘Fostering Patience in the Classroom: Results from Randomized
Educational Intervention’, Journal of Political Economy 126(5).
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’, Econo-
metrica 88(6).
Alibhai, S., Buehren, N., Frese, M. D., Goldstein, M. P., Papineni, S. and Wolf, K. (2019), ‘Full
Esteem Ahead ? Mindset-Oriented Business Training in Ethiopia’, World Bank - Policy Research
Working Paper Series 8892.
URL: https://ideas.repec.org/p/wbk/wbrwps/8892.html
Anderson, M. L. (2008), ‘Multiple inference and gender differences in the effects of early interven-
tion: A reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects’, Journal
of the American Statistical Association 103(484).
Andrews, J. L., Ahmed, S. P. and Blakemore, S. J. (2021), ‘Navigating the Social Environment in
Adolescence: The Role of Social Brain Development’, Biological Psychiatry 89(2).
URL: https://www.sciencedirect.com/science/article/abs/pii/S000632232031920X
Ardagna, S. and Lusardi, A. (2010), Explaining International Differences in Entrepreneurship: The

Role of Individual Characteristics and Regulatory Constraints, in J. Lerner and A. Schoar, eds,
‘International Differences in Entrepreneurship’, University of Chicago Press, chapter Chapter 1,
pp. 17 – 62.
URL: http://www.nber.org/chapters/c8224
Artinger, S., Vulkan, N. and Shem-Tov, Y. (2015), ‘Entrepreneurs’ negotiation behavior’, Small
Business Economics 44(4), 737–757.
URL: https://doi.org/10.1007/s11187-014-9619-8
Ashraf, N., Bau, N., Low, C. and McGinn, K. (2020), ‘Negotiating a better future: How
interpersonal skills facilitate intergenerational investment’, Quarterly Journal of Economics
135(2), 1095–1151.
32
Autor, D. H. (2015), ‘Why Are There Still So Many Jobs? The History and Future of Workplace
Automation’, Journal of Economic Perspectives 29(3), 3–30.
URL: https://www.aeaweb.org/articles?id=10.1257/jep.29.3.3
Bandiera, O., Buehren, N., Burgess, R., Goldstein, M., Gulesci, S., Rasul, I. and Sulaiman, M.
(2020), ‘Women’s Empowerment in Action: Evidence from a Randomized Control Trial in Africa’,
American Economic Journal: Applied Economics 12(1), 210–259.
Bandura, A. (2010), ‘Self-Efficacy’, The Corsini Encyclopedia of Psychology .

URL: https://doi.org/10.1002/9780470479216.corpsy0836
Bardasi, E., Gassier, M., Goldstein, M. and Holla, A. (2021), ‘The Profits of Wisdom: The Impact
of a Business Support Program in Tanzania’, The World Bank Economic Review 35(2).
Baumol, W. J. (1996), ‘Entrepreneurship: Productive, unproductive, and destructive’, Journal of

Business Venturing 11(1).
Bazerman, M. H., Curhan, J. R., Moore, D. A. and Valley, K. L. (2000), ‘Negotiation’, Annual
Review of Psychology 51(1), 279–314.
URL: https://doi.org/10.1146/annurev.psych.51.1.279
Bazerman, M. H. and Neale, M. A. (1994), Negotiating rationally, Free Press, New York.
Bertrand, M., Crépon, B., Marguerie, A. C. and Premand, P. (2021), ‘Do Workfare Programs Live
Up to Their Promises? Experimental Evidence from Cote D’Ivoire’, NBER Working Papers
Series w28664.
Blakemore, S.-J. (2012), ‘Imaging brain development: The adolescent brain’, NeuroImage 61(2).
URL: https://www.sciencedirect.com/science/article/abs/pii/S1053811911013620
Blakemore, S.-J. and Robbins, T. W. (2012), ‘Decision-making in the adolescent brain’, Nature
Neuroscience 15(9), 1184–1191.
URL: https://doi.org/10.1038/nn.3177
Blattman, C., Dercon, S. and Franklin, S. (2019), ‘Impacts of Industrial and Entrepreneurial Jobs
on Youth: 5-Year Experimental Evidence on Factory Job Offers and Cash Grants in Ethiopia’,
SSRN Electronic Journal .
Blattman, C., Fiala, N. and Martinez, S. (2020), ‘The Long-Term Impacts of Grants on Poverty:
Nine-Year Evidence from Uganda’s Youth Opportunities Program’, American Economic Review:
Insights 2(3).
Blattman, C., Jamison, J. C. and Sheridan, M. (2017), ‘Reducing Crime and Violence: Exper-
imental Evidence from Cognitive Behavioral Therapy in Liberia’, American Economic Review
107(4), 1165–1206.
URL: https://www.aeaweb.org/articles?id=10.1257/aer.20150503
33
Blattman, C. and Ralston, L. (2015), ‘Generating Employment in Poor and Fragile States: Evidence
from Labor Market and Entrepreneurship Programs’, SSRN 2622220.
Bloom, D., McKenna, M. and Prettner, K. (2018), ‘Demography, Unemployment, Automation,

and Digitalization: Implications for the Creation of (Decent) Jobs, 2010–2030’, NBER Working
Papers Series w24835.
Bontempo, R. and Iyengar, S. (2008), ‘Rio Copa: A negotiation simulation’, Columbia Caseworks,
New York .
URL: https://www8.gsb.columbia.edu/caseworks/node/96
Borghans, L., Ter Weel, B. and Weinberg, B. A. (2014), ‘People Skills and the Labor-Market
Outcomes of Underrepresented Groups’, ILR Review 67(2), 287–334.
URL: https://doi.org/10.1177/001979391406700202
Bowles, S., Gintis, H. and Osborne, M. (2001), ‘The Determinants of Earnings: A Behavioral
Approach’, Journal of Economic Literature 39(4), 1137–1176.
URL: https://www.aeaweb.org/articles?id=10.1257/jel.39.4.1137
Calderon, G., Cunha, J. M. and Giorgi, G. D. (2020), ‘Business Literacy and Development: Evi-
dence from a Randomized Controlled Trial in Rural Mexico’, Economic Development and Cultural
Change 68(2).
Campos, F., Frese, M., Goldstein, M., Iacovone, L., Johnson, H. C., McKenzie, D. and Mensmann,
M. (2017), ‘Teaching personal initiative beats traditional training in boosting small business in
West Africa’, Science 357(6357).
Card, D., Kluve, J. and Weber, A. (2018), ‘What Works? A Meta Analysis of Recent Active Labor
Market Program Evaluations’, Journal of the European Economic Association 16(3).
Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W. and Robins,
J. (2018), ‘Double/debiased machine learning for treatment and structural parameters’, Econo-
metrics Journal 21(1).
Choudhury, S., Blakemore, S. J. and Charman, T. (2006), ‘Social cognitive development during
adolescence’, Social Cognitive and Affective Neuroscience 1(3).
URL: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2555426/
Cialdini, R. (2001), ‘Harnessing the science of persuasion’, Harvard Business Review 79(9).
Cialdini, R. B. (2018), Influence: The Psychology of Persuasion, Revised Edition, Harper Collins
Publishers.
Dammert, A. C. and Nansamba, A. (2019), ‘Skills Training and Business Outcomes: Experimental
Evidence from Liberia’, Partnership for Economic Policy Working paper 2019-24.
34
de Mel, S., McKenzie, D. and Woodruff, C. (2013), ‘The Demand for, and Consequences of, Formal-
ization among Informal Firms in Sri Lanka’, American Economic Journal: Applied Economics
5(2), 122–150.
De Young, C. G. (2006), ‘Higher-order factors of the big five in a multi-informant sample’, Journal
of Personality and Social Psychology 91(6), 1138–1151.
Deming, D. J. (2017), ‘The growing importance of social skills in the labor market’, Quarterly
Journal of Economics 132(4).
Deming, D. and Kahn, L. B. (2018), ‘Skill requirements across firms and labor markets: Evidence
from job postings for professionals’, Journal of Labor Economics 36(S1), S337–S369.
URL: https://www.journals.uchicago.edu/doi/pdf/10.1086/694106
Dhaliwal, I., Duflo, E., Glennerster, R. and Tulloch, C. (2012), ‘Comparative Cost-Effectiveness
Analysis to Inform Policy in Developing Countries : A General Framework with Applications for
Education’, Abdul Latif Jameel Poverty Action Lab MIT.
URL: http://www.povertyactionlab.org/publication/cost-effectiveness
Dhar, D., Jain, T. and Jayachandran, S. (2020), ‘Reshaping Adolescents’ Gender Attitudes: Evi-
dence from a School-Based Experiment in India’, Northwestern Working Paper .
URL: https://www.nber.org/papers/w25331
Dumontheil, I. (2014), ‘Development of abstract thinking during childhood and adolescence: The
role of rostrolateral prefrontal cortex’, Developmental Cognitive Neuroscience 10.
Galinsky, A. D., Magee, J. C., Gruenfeld, D. H., Whitson, J. A. and Liljenquist, K. A. (2008),
‘Power reduces the press of the situation: Implications for creativity, conformity, and dissonance’,
Journal of Personality and Social Psychology 95(6).
Gennaioli, N., La Porta, R., Lopez-de Silanes, F. and Shleifer, A. (2013), ‘Human capital and
regional development’, Quarterly Journal of Economics 128(1).
Goldberg, L. R. (1993), ‘The Structure of Phenotypic Personality Traits’, American Psychologist

48(1).
Groh, M., McKenzie, D. and Vishwanath, T. (2015), ‘Labor contracts: Reducing information
asymmetries in the youth labor market of Jordan with psychometrics and skill based tests’,
World Bank Economic Review 29.
Heckman, 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).
Heller, S. B., Shah, A. K., Guryan, J., Ludwig, J., Mullainathan, S. and Pollack, H. A. (2017),
‘Thinking, Fast and Slow? Some Field Experiments to Reduce Crime and Dropout in Chicago*’,
35
The Quarterly Journal of Economics 132(1), 1–54.
URL: https://doi.org/10.1093/qje/qjw033
Hirsh, J. B., DeYoung, C. G. and Peterson, J. B. (2009), ‘Metatraits of the big five differentially
predict engagement and restraint of behavior’, Journal of Personality 77(4), 1085–1102.
International Trade Centre (2018), Promoting SME competitiveness in Africa data for de-risking
investment, International Trade Centre, World Trade Organization, and the United Nations.
Geneva, Switzerland.
URL: www.intracen.org/SMEBenchmarking
John, O., Donahue, E. and Kentle, R. (1991), ‘The big five inventory—Versions 4a and 54’, Journal
of Research in Personality 37(6).
Kautz, T., Heckman, J., Diris, R., ter Weel, B. and Borghans, L. (2014), ‘Fostering and Measur-
ing Skills: Improving Cognitive and Non-Cognitive Skills to Promote Lifetime Success’, OECD
Education Working Papers - OECD Publishing, Paris No 110.
Kerr, S. P., Kerr, W. and Xu, T. (2017), ‘Personality Traits of Entrepreneurs: A Review of Recent
Literature’, NBER Working Papers Series w24097.
URL: https://www.nber.org/papers/w24097
Latham, G. P. and Locke, E. A. (2007), ‘New developments in and directions for goal-setting
research’, European Psychologist 12(4), 290–300.
Levine, R. and Rubinstein, Y. (2017), ‘Smart and illicit: Who becomes an entrepreneur and do
they earn more?’, Quarterly Journal of Economics 132(2).
Lucas, R. E. (1978), ‘On the Size Distribution of Business Firms’, The Bell Journal of Economics
9(2).
McKenzie, D. (2020), ‘Small Business Training to Improve Management Practices in Developing

Countries: Reassessing the Evidence for Training Doesn’t Work’, Oxford Review of Economic
Policy Forthcoming.
McKenzie, D., Christopher, W., Bjorvatn, K., Miriam, B., Jing, C., Juanita, G.-U., Simon, Q.,
Tetsushi, S. and Martin, V. (2020), ‘Training entrepreneurs’, VoxDevLit 1(1).
URL: https://voxdev.org/lits/training-entrepreneurs
McKenzie, D. and Woodruff, C. (2017), ‘Business practices in small firms in developing countries’,
Management Science 63(9).
Murphy, K. M., Shleifer, A. and Vishny, R. W. (1991), ‘The Allocation of Talent: Implications for
Growth’, The Quarterly Journal of Economics 106(2).
36
OECD (2015), Skills for Social Progress: the power of Social and Emotional Skills, OECD.
URL: http://dx.doi.org/10.1787/9789264226159-en
Romano, J. P. and Wolf, M. (2005), ‘Stepwise multiple testing as formalized data snooping’, Econo-
metrica 73(4).
Schoar, A. (2010), ‘The Divide between Subsistence and Transformational Entrepreneurship’, In-
novation Policy and the Economy 10.
Schumpeter, J. A. (1911), The Theory of Economic Development, Harvard University Press, Cam-
bridge.
Sebastian, C., Viding, E., Williams, K. D. and Blakemore, S. J. (2010), ‘Social brain development
and the affective consequences of ostracism in adolescence’, Brain and Cognition 72(1).
URL: https://www.sciencedirect.com/science/article/abs/pii/S0278262609001055
Smith, A. (1776), An Inquiry into the Nature and Causes of the Wealth of Nations, Oxford Uni-
versity Press., London, England.
Thompson, L. (1990), ‘Negotiation behavior and outcomes: Empirical evidence and theoretical
issues’, Psychological Bulletin 108(3), 515–532.
URL: http://doi.apa.org/getdoi.cfm?doi=10.1037/0033-2909.108.3.515
Ubfal, D., Arraiz, I., Beuermann, D. W., Frese, M., Maffioli, A. and Verch, D. (2019), ‘The Impact
of Soft-Skills Training for Entrepreneurs in Jamaica’, SSRN Electronic Journal No. 12325.
URL: https://ssrn.com/abstract=3390314
van Dijke, M. and Poppe, M. (2006), ‘Striving for personal power as a basis for social power
dynamics’, European Journal of Social Psychology 36(4).
Yeager, D. S., Hanselman, P., Walton, G. M., Murray, J. S., Crosnoe, R., Muller, C., Tipton, E.,
Schneider, B., Hulleman, C. S., Hinojosa, C. P., Paunesku, D., Romero, C., Flint, K., Roberts,
A., Trott, J., Iachan, R., Buontempo, J., Yang, S. M., Carvalho, C. M., Hahn, P. R., Gopalan,
M., Mhatre, P., Ferguson, R., Duckworth, A. L. and Dweck, C. S. (2019), ‘A national experiment
reveals where a growth mindset improves achievement’, Nature 573(7774).
URL: https://www.nature.com/articles/s41586-019-1466-y.pdf
37

Soft Skills

Uploaded by

Copyright:

Available Formats

Soft Skills

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Soft Skills

Uploaded by

Copyright:

Available Formats

NBER WORKING PAPER SERIES

Working Paper 28845

NATIONAL BUREAU OF ECONOMIC RESEARCH

We gratefully acknowledge financial support from CEDIL/UKaid, ILO, Jacobs Foundation,

Laura Chioda Paul Gertler

3.1 Student Recruitment

3.2 Student Assignment to Treatment Arms and Training Sites

3.3 Teachers Training and Assignment

4 Data, Attrition, and Compliance

yi = α + βh Tih + βs Tis + γXi + i (1)

(1) (2) (3) (4) (5) (6) (7) (8)

Soft Skills Treatment

Mean of Hard Skills 0.761 0.465

Observations 1959 1959 1959 1959 1959 1959 1959 1959

5.1 Knowledge Assessments, Post Treatment

(a) Hard Skills Knowledge: Test Scores (z-score)

(b) Soft Skills Knowledge: Test Scores (z-score)

(1) (2) (3) (4) (5)

Hard skills Treatment

Soft skills Treatment

Mean control group 0.000 0.000 0.000 0.000 0.000

5.2 Hard and Soft Skills: 3.5 Years Post-Intervention

5.2.1 Task-Based Measures: Negotiation and Persuasion

(1) (2) (3) (4) (5)

Hard skills Treatment

Soft skills Treatment

Mean control group 0.814 321,017 3.749 0.022 0.026

5.3 Economic Outcomes

Table 4: Labor Market Outcomes: Employment Status

(1) (2) (3) (4)

Hard skills Treatment

Soft skills Treatment

Mean control group 0.672 0.431 0.336 0.095

(1) (2) (3) (4) (5)

Hard skills Treatment

Soft skills Treatment

Mean control group 0.474 0.725 0.336 1.810 1276

(a) log Total Current Earnings (b) log Wage Earnings

(c) log Profits (d) log Business Investments (Capital)

5.3.1 Total Earnings, Wage Earnings, and Profits

(1) (2) (3)

Hard skills Treatment

Soft skills Treatment

Mean control group (USD PPP adjusted) 229.58 195.83 6.25

5.4 Costs and Benefits

Table A.1: Baseline Balance Table: Baseline Sample

Control Hard Skills Soft Skills

Hard vs Soft vs Hard vs

Leadership index -0.03 0.00 0.04 0.345 0.071 0.349

Travel time (hours)

F-test of joint significance (F-stat) 1.27 1.20 0.59

Control Hard skills Soft skills

Hard vs Soft vs Hard vs

Leadership index -0.03 0.02 0.03 0.225 0.149 0.826

F-test of joint significance (F-stat) 1.73 1.18 1.22

Demographics and Expectations

Leadership index -0.010* (0.006)

Control Hard skills Soft skills

Soft vs Soft vs Hard vs

yi = α + βh Tih + βs Tis + γXi + i (1)