You are on page 1of 13

International Journal of Educational Development 71 (2019) 102089

Contents lists available at ScienceDirect

International Journal of Educational Development


journal homepage: www.elsevier.com/locate/ijedudev

Can scholarships provide equitable access to high-quality university T


education? Evidence from the Mastercard Foundation Scholars Program
Clemencia Cosentinoa, , Jane Fortsonb, Sarah Liuzzia, Anthony Harrisc, Randall Blaird

a
Mathematica, 1100 1st Street, N.E., Washington, DC, USA
b
Mathematica, 2101 4th Avenue, Suite 1350, Seattle, WA, 98121, USA
c
Mathematica, 955 Massachusetts Avenue, Suite 801, Cambridge, MA, 02139, USA
d
Mathematica, 111 East Wacker Drive, Suite 920, Chicago, IL, 60601, USA

ARTICLE INFO ABSTRACT

Keywords: This paper estimates the impact of a scholarship program—the Mastercard Foundation Scholars Program—on
Higher education access to university. Using propensity score matching, the analysis finds that the Program improves access to
Human capital high-quality university studies among vulnerable, disadvantaged youth, mostly from Africa—increasing timely
Access to education access to university, to higher levels of study, and to studies abroad. Half of scholarship recipients studying
Scholarships
abroad return to Africa and a third pursue graduate studies within 18 months of graduation. The Program’s
Leadership development
collaborative approach holds promise for scaling up this and similar efforts to increase university participation
among youth in African and other nations.

1. Introduction recent World Bank report shows that tertiary enrollment has grown
more for youth in households in the top 20 percent of the income dis-
Insufficient and inequitable access to high quality tertiary education tribution (for whom there has been a 7.9 percent increase between
stands in the way of economic development and poverty eradication in 1998 and 2012) than for those in the bottom 80 percent (increase of 3.1
Sub-Saharan Africa (SSA). Tertiary education fuels modern economies; percent over the same period) (Darvas et al., 2017). After conducting an
nations in SSA need a skilled workforce to meet employer needs and extensive empirical analysis of equity in tertiary enrollment in SSA
help their economies grow and diversify. Tertiary education is also countries, the authors of this report reinforce what others had found in
associated with individual benefits in the form of higher earnings, the past, namely, that “access to tertiary education in [SSA] has unduly
particularly among graduates in SSA (compared to other regions of the benefited students drawn from the region’s wealthiest households, and
world) and among women (compared to men) (Montenegro and overall enrollment remains disproportionately male and metropolitan.
Patrinos, 2014; Darvas et al., 2017). This suggests a strong connection These factors stifle the catalytic potential of higher education….Instead,
between tertiary education and improved well-being. Yet, at present, patterns of access to tertiary education have generally reinforced and
SSA has the lowest average tertiary gross enrollment ratio in the world reproduced social inequality” (Darvas et al., 2017).
(9 percent in 2017, compared to 38 percent globally) and the fastest The Mastercard Foundation Scholars Program was created in 2012
pace of population growth (UIS, 2017; Canning, 2015; UN, n.d.). Ter- to help address insufficient and inequitable access to high quality ter-
tiary enrollment has been growing faster in SSA than in the rest of the tiary education. By design, the Scholars Program aims to serve youth
world, absorbed by tremendous expansion in public and private tertiary with demonstrated leadership potential who have traditionally been
institutions (UIS, 2017 and Darvas et al., 2017, though their estimates overlooked by educational systems. These include girls and young
differ). But this growth has been insufficient to match rising demand for women, and economically disadvantaged, displaced, and rural youth.
tertiary education, which is driven by factors such as rapid population The Program—which offers comprehensive scholarships to youth pri-
growth and improved participation in primary and secondary education marily from Africa—seeks to identify and educate next-generation
(UIS, 2010; Darvas et al., 2017). leaders that come from these disadvantaged, underserved groups to
In addition to shortfalls in the supply of tertiary education oppor- help drive a process of social and economic transformation in their
tunities, access to those opportunities remains deeply inequitable. A communities, countries, and, ultimately, the African continent.

Corresponding author.

E-mail addresses: ccosentino@mathematica-mpr.com (C. Cosentino), jfortson@mathematica-mpr.com (J. Fortson), sliuzzi@mathematica-mpr.com (S. Liuzzi),
aharris@mathematica-mpr.com (A. Harris), rblair@mathematica-mpr.com (R. Blair).

https://doi.org/10.1016/j.ijedudev.2019.102089
Received 3 March 2019; Received in revised form 7 June 2019; Accepted 21 July 2019
Available online 25 September 2019
0738-0593/ © 2019 Mastercard Foundation. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

Scholarship programs are a longstanding human capital develop- the longest standing international scholarship program. More recent
ment strategy, particularly in developing countries. The latest United programs include the Fulbright Scholarships established by the federal
Nations’ Sustainable Development Goals include, as a target for 2020, government of the United States to promote international exchanges
to “substantially expand globally the number of scholarships available and understanding (1946), the Commonwealth Scholarship and Fel-
to developing countries, in particular least developed countries, small lowship Plan through which member governments of the Common-
island developing States and African countries, for enrolment in higher wealth offer scholarships and fellowships to citizens of other Com-
education.” Governments, private philanthropies, and other organiza- monwealth countries (1959), and the Ford Foundation International
tions are undertaking efforts to achieve this goal. For example, the Fellowship Program for emerging leaders in the developing world to
Chinese government has been making progress towards this goal since pursue graduate studies (2001).
2006 by pledging scholarship support to African students wishing to As these examples suggest, scholarships vary along many dimen-
study in China, most recently promising to provide 30,000 scholarships sions, such as the organization that creates them (government, private
by 2018 (Breeze and Moore, 2017; Makundi et al., 2017). philanthropy, or multi-lateral organization), the goals of the scholar-
The Mastercard Foundation Scholars Program stands out not only ship (to foster cultural understanding, forge ties with other nations,
for its programmatic approach—which offers participating students promote human capital development, and so on), the education level
(“Scholars”) a host of support services to complement the financial (undergraduate, graduate, or post-graduate study) or experience sup-
support—but also by providing a unique opportunity to advance ported (studies, research, or professional experience), the population
knowledge that can catalyze broader changes. Most scholarship pro- targeted (such as nationals of specific nations, vulnerable youth, or
grams serve limited numbers of youth in any given year and rarely students seeking careers in specific fields); the duration of the scho-
devote substantial funding to research and evaluation activities. By larship (one year, multiple years, and so on); whether they provide full
serving large numbers of youth and by supporting research and eva- versus partial funding; whether they require beneficiaries to return
luation studies since the inception of the Program, the Scholars home; and other factors.
Program can help fill critical knowledge gaps. Research and evaluation
activities supported by the Scholars Program not only improve the 2.1. Literature review
Program, but they also inform public and private sector efforts to cat-
alyze positive change through tertiary education and leadership de- Although a comprehensive study of existing international scholar-
velopment. This article shares some of the first findings from research ship programs has yet to be conducted, Perna and colleagues (2014)
and evaluation activities sponsored by the Scholars Program. completed a review of government-sponsored international scholar-
Despite the longstanding and widespread use of scholarships in ships. They identified 183 international scholarship programs in 196
public and philanthropic efforts to increase participation in tertiary nations. Many others are sponsored by private philanthropies—such as
education and promote social mobility, there is very limited evidence the Mastercard, Ford, and Bill & Melinda Gates foundations. What is
about their impact (Mawer, 2017; Martel, 2018). The Scholars Program striking, given the large number of scholarship programs, is the dearth
is anchored on a theory of change that aligns with policies used today of evidence regarding their effectiveness.
throughout the world to improve participation and equity in tertiary To identify current practices and assess the potential for more rig-
education, but often without solid evidence of their effectiveness. These orous evaluation, the Commonwealth Scholarship Commission in the
include encouraging universities to seek out promising students that UK commissioned a study of research methodologies used to evaluate
would otherwise not be targeted by recruitment efforts, using socio- scholarship programs. Findings indicate that counterfactual designs are
economic characteristics of applicants for admissions decisions, and rare and “almost all evaluations are ex-post,” retrospective efforts to
enhancing financial aid offerings and wraparound support for eco- identify former recipients and survey them to learn about their ex-
nomically-disadvantaged students. These strategies constitute a proac- periences and outcomes (Mawer, 2014). In fact, the author finds that
tive, intentional effort to improve access by (1) encouraging revisions to nearly all scholarship programs reviewed conducted this type of “tracer
recruitment policies and practices used by institutions of higher edu- study,” and that “Randomised Control Trials (RCTs), regression dis-
cation and (2) addressing financial and information barriers faced by continuity design, and difference-in-differences designs, amongst other
underserved youth. Using rigorous methods to study the impact of the possible [rigorous] approaches, have not been reported within the
Scholars Program on access to higher education, this study contributes sector.”
evidence on the effectiveness of these strategies to catalyze transfor- Our own review of the literature confirms this finding. There is a
mative social and economic change, evidence that has important im- large body of research showing that scholarship programs in developing
plications for institutions of higher education, public policy, and phi- countries can improve access and attainment for economically dis-
lanthropic/private sector efforts to leverage the demographic changes advantaged groups, but these studies largely focus on secondary edu-
sweeping the African continent. cation (Filmer and Schady, 2009; Angrist et al., 2006; Barrera-Osorio
Below, we review the existing literature to contextualize this study et al., 2007; Behrman et al., 2005; Al-Samarri and Zaman, 2006;
and describe the approach developed by Mastercard Foundation to ef- Deininger, 2003; Meng and Ryan, 2010; Duflo et al., 2015; Attanasio
fect change (Section 2), describe our impact and descriptive analyses et al., 2010; and Gajigo, 2016). A handful of studies have collected
(Section 3), present findings (Section 4), and discuss the implications of evidence of effectiveness among scholarship programs that are similar
those findings (Section 5). to the Scholars Program in targeting disadvantaged and underserved
youth and offering wrap-around services in addition to financial sup-
port, but for students who are United States citizens or permanent re-
2. Background sidents.
For example, the evaluation of the Gates Millennium Scholars
Scholarships for international students pursuing university studies (GMS) Program, which is a scholarship program aimed at low-income
have been available for decades and are remarkably diverse.1 Estab- minority students from the U.S., used propensity score matching to
lished in the early 1900s, the prestigious Rhodes Scholarship is likely create a comparable group of nonparticipants. Findings indicate that,
for some subgroups of students, the GMS Program improved (1) access
1
As we know them today, scholarships date back about a hundred years. to highly selective institutions of higher education, (2) retention in and
However, Perna and colleagues (2014) note that similar schemes date back to completion of postsecondary degrees, and (3) enrollment in post-bac-
the colonial period when nations offered study abroad programs to train the calaureate and graduate degree programs (Amos et al., 2009). How-
administrative elites for their colonies. ever, these findings were not consistent across cohorts, gender, or racial

2
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

groups and the estimated differences, when present, were small (3 to 4


percentage-point difference in retention or being “on track academi-
cally” between participants and nonparticipants, for example).
The evaluation of the National Science Foundation Louis Stokes
Alliances for Minority Participation program, which offers scholarships
and support services to students from underrepresented racial and
ethnic backgrounds in science, technology, engineering, and mathe-
matics (STEM), focused on access to graduate school, a key program
goal. Findings show that, on average, program participants were more
likely to pursue additional post-bachelor’s coursework, enroll in grad-
uate programs, and complete advanced degrees, particularly in STEM,
than did national comparison groups (Clewell et al., 2006). The dif-
ferences found were large; for example, the study found that 65 percent
of program participants went on to graduate school, compared to 45
percent in the comparison samples. Our study also used a quasi-ex-
perimental design to contribute to this literature by focusing on a si-
milar but international scholarship program—the Mastercard Founda-
tion Scholars Program—the largest privately-sponsored scholarship
program for Africans—to study its impact on access to university.

2.2. The Scholars Program

The Scholars Program is anchored on a belief in “the transformative


power of individuals as catalysts for broader change” (Burciul and Kerr,
2018). To enable individual agency, the Program aims to expand access
to high quality education for youth who exhibit academic promise,
economic disadvantage or vulnerability, leadership potential, and a
Fig. 1. Study Partner Institutions at Tertiary Level.
commitment to giving back to society.
The Program includes financial and non-financial supports to help
Scholars succeed in their studies and transition to opportunities after students comprising around 84 percent of tertiary Scholars. As of early
graduation. These include academic and social supports (such as or- 2019, the program has committed nearly $850 million to offer 9,000
ientation and tutoring); leadership, entrepreneurship, and community university and 26,000 secondary education scholarships.
service training activities; and access to networks, resources (such as
counseling), and opportunities for professional preparation (such as 3. Data and methods
internships, mainly in Africa to strengthen ties to the continent among
students studying abroad). This holistic approach is designed to prepare Using data on early cohorts of tertiary Scholars, we aim to (1) es-
and empower Scholars to become ethical leaders and promote positive timate the impacts of the Program on access to undergraduate educa-
change in their communities. tion and (2) describe the post-Program outcomes of early Scholars
To offer supports to Scholars, the Foundation formed a global Program alumni. The analysis is based on a survey of Scholars and
partnership with non-governmental organizations (NGOs) and a lea- comparison applicants who had applied for the Scholars Program for
dership academy in Africa, and some of the world’s leading universities undergraduate studies (impact analysis) and a survey of undergraduate
in Africa, Europe, Central and North America, and the Middle East. The and graduate Program alumni administered after graduation (de-
Scholars Program has operated at both the secondary and tertiary le- scriptive analysis). Our impact and descriptive approaches and data
vels, and has worked with different partners over time; our analysis sources are explained in greater detail in the sections that follow.
focuses on the tertiary level (Fig. 1 shows tertiary partners included in
this study). 3.1. Impact analysis
Although the Foundation established broad qualifying criteria for
the Scholars Program, partners are responsible for recruiting and se- 3.1.1. Data
lecting Scholars and supporting them throughout their studies. Most The impact analysis focuses on the tertiary program at the under-
partners rely on program advertisements (online and through media, graduate level. The data used in the analysis come from a survey of
public events, and job fairs) to generate interest in the Program, though applicants to undergraduate programs who were also considered for the
a few partners target their recruitment efforts with secondary-level Scholars Program. These applicants sought admission to begin their
Scholars Program partners or feeder schools. At the selection stage, university studies between fall 2013 and spring 2015 and join the
partners screen candidates’ application materials to assess their aca- Scholars Program’s 2014 or 2015 cohorts. (The universities partici-
demic admissibility, economic need, and social vulnerability and then pating in the Program are spread across continents and have start dates
make final selections after conducting interviews or home visits with throughout the year.) We obtained the data to identify and contact
short-listed candidates. To participate in the Scholars Program, appli- applicants from 18 institutional Program partners, mostly universities
cants must meet university admissions requirements in addition to in North America and Africa. (The partners are listed in Fig. 1 above;
those set by the Program. Those selected for the Program receive the Appendix Table A1 shows the universities attended by Scholars.) We
academic and social supports described above as well as a financial collected information on applicants to ensure we surveyed both those
package that varies by partner but generally includes tuition and fees; selected for the Scholars Program and those who were not selected
books, a laptop, and other learning materials; accommodation; a living (comparison applicants or the comparison group).
stipend; health insurance and other wellbeing supports; and travel ex- Overall, 72 percent of the study population responded to the survey,
penses. 91 percent of Scholars and 56 percent of comparison applicants
Since its inception in 2012, the Program has served nearly 5,100 (Appendix Table A2). The number of respondents in the comparison
undergraduate and nearly 1,000 graduate Scholars, with undergraduate group provided sufficient data to create a matched comparison group

3
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

for analysis (described in Section 3.1.2). The final analysis is based on partners, respondents from the 2015 cohort who applied to participate
1,147 completed surveys, 637 from Scholars and 510 from matched in the Program through Camfed, and respondents from the 2015 cohort
comparison applicants. who applied to partners outside of North America other than Camfed.
The survey included questions about applicants’ family background, (Camfed is a Program partner that provides support to students who
experiences during secondary school, and enrollment in university, enroll at universities throughout Ghana.) This approach accounts for
among others. Administered in 2014–2015, the survey coincided with some of the differences in the application process across cohorts and
the first year of expected university studies for the 2015 cohort and the partners by only matching comparison applicants with Scholars in the
second year for the 2014 cohort. The timing of the survey enabled us to same application group. Camfed was the only partner with a large en-
capture actual enrollment in university, not just admission, for two ough applicant pool to serve as its own comparison group. To ensure
cohorts of applicants. that Scholars and their matched comparisons were indeed similar along
key dimensions, we restricted the matching algorithm to generate
3.1.2. Propensity score matching approach matches only if comparison applicants were sufficiently close to the
To assess the impact of the Scholars Program on access to higher Scholars—defined as having a propensity score within a particular
education, we used a quasi-experimental design with propensity score bandwidth of the propensity score (in this case, 0.06). We found sui-
matching methods to create a comparison group that was ob- table matches for 89 percent of Scholars.
servationally similar to the Scholars group, and therefore indicative of Table 1 displays the characteristics of Scholars and comparison
what would have happened to the Scholars if they had not participated applicants in the full sample and in the matched sample used for ana-
in the Program. The comparison group is composed of similar appli- lysis. It shows that matching yields two groups (Scholars and compar-
cants who were not selected for the Program. We matched Scholars to ison) that are observationally similar at baseline. In other words, the
comparison applicants through logistic regression models that account analysis sample represents an improvement over using the full (un-
for characteristics related to the likelihood of being selected into the matched) sample. At baseline, there were significant average differ-
Program, including academic achievement, past participation in lea- ences between the Scholars and comparison groups in most character-
dership activities, and family socioeconomic characteristics. istics—such as gender, academic achievement, and socioeconomic
The matching process was conducted separately within four groups: characteristics—and nearly all of those differences disappear after
respondents from the 2014 cohort across partners/institutions, re- matching.
spondents from the 2015 cohort who applied to North American

Table 1
Differences in characteristics of Scholars and comparison groups before and after propensity score matching.
Baseline characteristics Before matching After matching
(full sample) (matched sample)

Difference between Scholars and Average for Difference between Scholars and Average for
comparison applicants Scholars comparison applicants Scholars

Demographic characteristics
Male −8.8** 37.0% 2.2 37.4%
Age 0.4* 21.0 0.1 20.9

Leadership, volunteering and academic achievement


Participated in community activities in secondary 1.9 93.9% 1.3 93.5%
school
Participated in leadership activities in secondary 6.2** 69.1% 2.7 69.4%
school
Participated in service activities in secondary 1.1 94.8% 0.7 94.6%
school
Ranked among top 10 percent of students in year 7.5** 50.4% 1.1 51.5%
in secondary schoola
Does not know rank −9.0** 25.3% 0.9 24.6%
Took standardized test before applying 9.0** 31.6% 0.0 30.9%

Secondary school characteristics


Attended public secondary school 7.9** 71.6% 1.6 72.3%
Attended boarding school 10.4** 71.2% 2.1 70.0%

Socioeconomic status
Lived in rural area 1.7 29.2% 2.1 27.1%
Home had unfinished floor 7.8** 23.5% 3.0 21.8%
Home had no electricity 0.6 23.1% 4.2* 21.8%
Home had no flush toilet 8.4** 69.2% 0.4 67.6%
Home had no stove 11.8** 69.8% −0.1 68.7%

Family situation and parental education


Lived with guardians 2.9 18.9% 0.8 18.7%
Mother has at least some secondary educationa −9.0** 36.1% −0.6 37.2%
Does not know mother's education 0.2 13.4% 1.6 13.5%
Father has at least some secondary educationa −8.0** 43.8% −0.2 44.3%
Does not know father's education 1.8 22.0% 0.5 22.3%

Notes: Table shows differences in pre-university/baseline characteristics between Scholars and comparison applicants. Averages and differences in the “before
matching” column are calculated using data for 687 Scholars and 515 comparison applicants. Averages and differences in the “after matching” column are calculated
using 637 Scholars and 510 comparison applicants, in addition to applying the weights derived from the propensity score matching process. The number of Scholars
and comparison applicants decreases due to the exclusion of respondents that are outside of common support.
*Significantly different from zero at the 0.05 level. **Significantly different from zero at the 0.01 level.
a
Reported percentage among all respondents, even if the respondent reports “don’t know”.

4
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

3.1.3. Analytic approach responses to questions about Scholars’ education and employment after
After matching, we estimate the impact of the Program as the dif- graduation, including whether they remain in (if studying at home) or
ference between the average outcomes of Scholars and the matched return to (if studying abroad) their home countries. We report de-
comparison group. We use ordinary least squares regression models to scriptive statistics of outcomes of interest, such as the share of alumni
estimate the Program’s effect (this difference) on three binary out- entering the labor force and pursuing further education. All estimates
comes: enrolling in a university, enrolling in a bachelor’s degree pro- are weighted to adjust for non-response (as described in the Appendix).
gram (or equivalent), and enrolling in a university abroad. The basic
equation to estimate impacts takes the following form: 4. Results
Yim = + Sim + Xim + im (1)
The Scholars Program targets high-achieving youth with demon-
where Yim is the outcome of applicant i in the matched sample m . Sim is strated leadership potential who come from disadvantaged, under-
an indicator equal to 1 if applicant i became a Scholar and equal to zero served groups. The characteristics of the Scholars included in the ana-
otherwise, Xim is a vector of controls for individual applicants’ baseline lysis reflect these aims. Of Scholars included in the impact analysis,
characteristics, and im is a random error term. The coefficient signals three-quarters are female, about one-third are from rural areas, and the
the Program impact. Comparison applicants who are more similar to majority have at least one parent with no more than primary education
Scholars—and therefore have more similar propensity scores—are (Table 1). Scholars performed well in secondary school; of those who
given more weight in the estimation using kernel weights. know their rank, two-thirds report that they were among the top 10
In Section 4 of this paper, we present regression-adjusted estimates percent of students in their year in secondary school, and also report
obtained through models that predict the three university enrollment high levels of community engagement, leadership, and service during
outcomes as a function of participation in the Scholars Program, con- their secondary school studies (Table 1). In the sections that follow, we
trolling for baseline characteristics used in the propensity score discuss how the Scholars Program affects access to tertiary education
matching models. Using baseline characteristics as control variables in (Section 4.1) and describe the education and work experiences of
the models increases the precision with which we estimate effects and Scholars after graduation (Section 4.2)
helps adjust for any remaining differences between the Scholar and
comparison groups. We estimate impacts overall and by cohort because 4.1. Impact of the Scholars Program on access to university education
comparison applicants in the 2014 cohort had more time to enroll in
university than those in the 2015 cohort, potentially influencing our Results from the impact analysis show that the Mastercard
impact estimates. Furthermore, the composition of the study sample Foundation Scholars Program increases access to a university educa-
differed across cohorts (as described below). tion. The Program has a large and positive effect on all access-related
outcomes studied: enrollment in university, in a bachelor’s degree
3.2. Descriptive analysis program, and in a university abroad. Compared to applicants who were
not selected for the Scholars Program, Scholars are 30 percentage
The data used in the descriptive outcomes analysis come from a points more likely to enroll in a university, 37 percentage points more
2017 survey of Scholars Program alumni who graduated between June likely to enroll in a bachelor’s degree program, and 33 percentage
2015 and July 2016. At the time of the survey, an average of 1.5 years points more likely to enroll in a university outside their home country
had elapsed since Scholars’ graduation, giving them time to pursue (Table 2). This indicates that the Program succeeded in its goal of im-
further education or join the workforce. A total of 155 tertiary-level proving access to high quality education for high-achieving students
Program alumni responded to the survey, or 69 percent of the 225 who might have otherwise not pursued post-secondary education.
Scholars Program alumni served through 12 Program partners who These overall impacts varied by cohort (Table 3). Scholars in the
graduated from 18 colleges and universities in Africa, Central America, 2015 cohort were 33 percentage points more likely than comparison
North America, and the Middle East. Our analytic sample includes 124 applicants in that cohort to enroll in university—nearly all Scholars in
alumni who completed all or most of the survey. (The partners are listed the 2015 cohort were enrolled, compared to two-thirds of comparison
in Fig. 1; Appendix Table A1 indicates the universities attended by applicants. In the 2014 cohort, the impact was much smaller (9 per-
alumni.) centage points), in part because some Scholars in that cohort were no
For the descriptive analysis of outcomes included here, we relied on longer enrolled, but largely because comparison applicants from the

Table 2
Impacts of the Scholars Program on university enrollment (regression estimates using matched sample).
Outcome Overall impact on enrollment Average enrollment for Scholars Scholars (N) Comparison (N)
(percentage points)

Currently enrolled in university 29.5** 98.4%a 634 508


Currently enrolled in Bachelor’s (or equivalent) program 37.4** 93.7%b 573 496
Currently enrolled in university outside home country 33.3** 44.1% 629 499

Notes: Impact estimates (shown in the first column) are calculated using a propensity score matching model that uses kernel-density matching methods. We estimate
robust standard errors; statistical significance (difference from zero) is denoted by: ** p < 0.01, * p < 0.05. Sample sizes vary due to missing data. The results
reported are from regressions that control for gender, age, indicators for whether mother or father attained secondary education, an indicator for living with parents,
attendance at public secondary school, attendance at boarding secondary school, participation in community, leadership, and service activities in secondary school,
ranking in the top ten percent of class, an indicator for taking a standardized test, rural/urban status, and housing characteristics. The regressions also control for the
broad groupings of Mastercard Foundation partners (such as by geography) that participated in the study and interactions between these groupings and the control
variables.
a
This estimate is not 100% because there are 10 Scholars who responded to the survey and indicated that they were no longer enrolled at the time of the survey.
This group likely reflects a small number of Scholars who have dropped out of their universities and a small number of Scholars who were enrolled but stated
otherwise (for instance, because they completed the survey during a school break).
b
Scholars who are enrolled in university—but not enrolled in Bachelor’s programs—are enrolled in technical and certificate programs, including in nursing,
midwifery, and education.

5
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

Table 3
Impacts of the Scholars Program on university enrollment, by cohort (regression estimates using matched sample).
Outcome and sample Overall impact on enrollment (percentage points) Average enrollment for Scholars Scholars (N) Comparison (N)

Currently enrolled in university


2014 cohort 9.4* 94.3% 140 130
2015 cohort 32.9** 99.5% 494 378
2014 and 2015 cohorts 29.5** 98.4% 634 508

Currently enrolled in Bachelor’s or equivalent


2014 cohort 12.1* 92.9% 127 123
2015 cohort 43.4* 93.9% 446 373
2014 and 2015 cohorts 37.4** 93.7% 573 496

Currently enrolled in university outside home country


2014 cohort 54.6** 66.4% 140 129
2015 cohort 27.6** 37.8% 489 370
2014 and 2015 cohorts 33.3** 44.1% 629 499

Notes: See notes in Table 2.

2014 cohort were more likely to be enrolled (85 percent) relative to percentage points more likely than the comparison group to be en-
comparison applicants from the 2015 cohort.2 rolled. While still large and positive, this impact is smaller (25 per-
There are several possible explanations for the differences between centage points) among applicants to non-North American partners. This
cohorts. First, impacts on university enrollment might be smaller for the likely signals the difficulties that the target population—high achieving
2014 cohort because unsuccessful applicants from the 2014 cohort had youth from disadvantaged socio-economic backgrounds—faces in ac-
time to reapply and gain admission to university and obtain funding, cessing university studies abroad. It also aligns with the earlier finding
whereas applicants from the 2015 cohort had not yet had as much time that the Program has a large and positive impact on enrollment in a
to do so. Consistent with this hypothesis, many of the comparison ap- university abroad.
plicants in the 2014 cohort started university a year later than Scholars To better understand the findings by region and cohort, we ex-
in the same cohort. amined the institutions attended by comparison applicants. Many re-
Second, impacts might be larger for the 2015 cohort because partner spondents who applied to non-North American partners but did not
institutions might have improved their recruitment practices, allowing receive scholarships are enrolled at a university or college, usually in
them to better identify youth who would not attend university without their home country. In the case of Ashesi University, KNUST, and
the Scholars Program. Through interviews with Program partners in Makerere University, many comparison applicants not selected for the
2014 and 2016, we learned that partners initially struggled to recruit Scholars Program nevertheless enroll at the institution to which they
candidates that met Program qualifications (for instance, because they applied (Ashesi, KNUST, or Makerere), but not through the Scholars
lacked the student pipelines or the experience recruiting these candi- Program. In contrast, comparison applicants who applied to North
dates, or because qualifying students had difficulty navigating or American partner institutions are either not enrolled at the time of the
paying for the application process). Over time, however, partners survey, enrolled at another North American university, or enrolled
strengthened their recruitment practices (for example, by building elsewhere. This difference across applicants to North American and
partnerships with local organizations and collaborating with secondary- non-North American partners is consistent with the hypothesis that
level Program partners in Africa) to enhance their ability to attract some unsuccessful applicants to North American universities are
applicants that met Program admissions criteria and help youth better choosing to wait a year to try and secure funding to attend university
understand the application process. To the extent that such strategies abroad.
started bearing fruit from one year to the next, this might help (par-
tially) explain the larger observed impacts in the 2015 cohort.
4.2. Post-graduation outcomes of the Scholars Program
Lastly, the difference in impacts between cohorts could reflect the
composition of the study sample, which changed between 2014 and
The findings in the prior section demonstrate that the Scholars
2015 when new institutions joined the Program (and study) and an
Program has large and positive impacts on access to higher education. It
existing partner joined the study. The 2014 cohort has sizeable samples
will take several years to assess whether the Program also has an impact
from partners outside of Africa, including EARTH University (Costa
on other outcomes, such as further education. To get a preview of such
Rica), the African Leadership Academy (whose students largely attend
outcomes, we surveyed earlier cohorts of Program alumni in 2017, or
universities abroad), and Michigan State University (USA). Several
about 1.5 years (on average) after graduating. The analysis sample in-
partner institutions in Africa joined the study in the 2015 cohort
cludes 54 Scholars who had completed undergraduate degrees and 70
(KNUST, Makerere University, and the University of Cape Town, and
Scholars who had completed graduate degrees through the Scholars
Ghanaian universities attended by Scholars supported by Camfed).
Program.
Partners vary in their admissions criteria and the characteristics of their
These Scholars were excluded from the impact analysis, as they had
applicants, on average. Compositional differences could explain some
either entered the program as part of the 2013 Program cohort or in
of the differences across cohorts.
graduate level programs of study, which were not included in the im-
Program impacts also vary by the partner’s region (Table 4). The
pact study. Surveying them enabled us to learn about several of the
Program has a larger impact on access to university for applicants to
critical expected outcomes given the Program’s theory of change—in-
North American partners (Canada and the United States) than partners
cluding whether Scholars returned to, or stayed on, the African con-
in other regions of the world, namely, Africa, Central America, and the
tinent after graduating, pursued higher level graduate studies (master’s
Middle East. Scholars who applied to North American partners are 42
and doctoral degrees), or entered the labor force (through the formal
sector or entrepreneurship). These findings apply to the particular co-
2
The comparison applicant enrollment rates presented here (85 percent for hort included in the study; we might expect outcomes to be different for
the 2014 cohort and 67 percent for the 2015 cohort) reflect the enrollment rate later cohorts as the composition of the partnership has changed and the
among comparison applicants in the analysis sample after weighting. Program—as implemented by different partners—also evolved.

6
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

Table 4
Impacts of the Scholars Program on university enrollment, by partner’s geographic region (regression estimates using matched sample).
Outcome and sample Overall impact on enrollment (percentage points) Average enrollment for Scholars Scholars (N) Comparison (N)

Currently enrolled in university


Applicants to North American partners 41.7** 98.8% 180 94
Applicants to Non-North American partners 25.2** 98.2% 454 414
All applicants 29.5** 98.4% 634 508

Currently enrolled in Bachelor’s or equivalent program


Applicants to North American partners 43.0** 97.6% 168 93
Applicants to Non-North American partners 35.5** 92.1% 405 403
All applicants 37.4** 93.7% 573 496

Notes: See notes in Table 2. Applicants to the African Leadership Academy—which places students in both North American and non-North American universities—are
included with North American partners.

Nevertheless, these findings are indicative of whether the Program is on with most being accepted. In contrast, of the 70% of undergraduate
track to achieve its goals. alumni currently working, around three-quarters report having salaried
Although tertiary alumni come from more than a dozen countries, positions in full-time employment and less than one-third reported
over half of alumni come from just four countries (Ghana, Kenya, applying to a university program in the past 6 months.
Lebanon, and Uganda).3 Two-thirds of alumni studied in their home An important goal of the Scholars Program is to develop transfor-
country. Nearly all of the African Scholars who studied in Africa re- mative leadership—among individuals (by supporting the education of
mained on the continent, and about two thirds of those alumni re- underserved and vulnerable groups) and institutions of higher educa-
mained in or returned to their home countries (Table 5). tion (topic not covered in this paper). Our findings suggest that women
Among African alumni who studied abroad, about half returned to are more likely than men to join the labor force and less likely than men
the African continent, with most returning to their home country. to enroll in further graduate-level university studies (Table 6). We ob-
Among those who studied abroad and had not returned to Africa at the serve a similar pattern for youth who grew up without either parent:
time of the survey (that is, within about 1.5 years of graduation), less they are more likely to be employed and less likely to pursue further
than 20 percent indicated that they intend to return home in the near graduate studies (estimates not shown). Family and financial respon-
future. sibilities may partially explain these findings.
Most Scholars who had graduated by 2017 (56%) had earned
graduate degrees, primarily in mathematics, statistics and science 5. Discussion
fields. Others earned undergraduate degrees in a wide range of fields,
particularly life and physical sciences, as well as engineering.4 At the The findings from this study provide solid evidence that a long-
time of the survey, the majority of alumni were working (about 60%) or
standing practice—offering scholarships—is an effective way to im-
pursuing further studies (over 30%) (Table 6). prove access to university studies among vulnerable, disadvantaged,
Private businesses (34%) and schools or universities (25%) provided
and underserved youth that constitute the large majority of Mastercard
most jobs to these graduates, followed by the non-profit (16%) and Foundation Scholars. In particular, the findings suggest that scholar-
public or governmental (12%) sectors. A small share of alumni became
ships increase timely access to higher education, to higher levels of
entrepreneurs (less than 10%). Those pursuing further education pri- study (such as bachelor’s degree programs instead of certificate pro-
marily enrolled in master’s or honors degrees (73%), mostly in
grams), and to studies abroad. We find that the impact on enrollment in
mathematics, statistics, agriculture, or health. Scholars’ career trajec- university is smaller for the earlier cohort (2014) in our study, which
tories differ, however, by type of degree and individual characteristics
could indicate that the enrollment gap could narrow over time. We also
(Table 6). In this cohort, undergraduate alumni are more likely than find larger impacts on applicants to North American universities
graduate alumni to be working—instead of pursuing further stu-
(compared to applicants to universities elsewhere), which could in-
dies—about 1.5 years after graduation. Interestingly, both graduate and dicate that unsuccessful applicants to these institutions defer college-
undergraduate alumni who are not studying reported financial con-
going for some time. Future follow-up with these (and other) cohorts
straints as a reason—although undergraduate alumni were more likely would allow us to estimate the persistence of the estimated impacts, in
to express a desire to join the workforce as the reason they are not
addition to capturing a broad range of relevant outcomes, such as
currently studying. graduation, transition to future education and the workforce, and lea-
If they do continue their studies, undergraduate alumni over-
dership and give-back.
whelmingly enroll in honors and master’s programs. In contrast, alumni All nations—not just those in SSA—may find this evidence useful as
who earned graduate degrees are more likely to continue studying,
they consider ways to improve the life chances of those groups that are
enroll in PhD programs, and rely on scholarships as their primary not well represented among postsecondary graduates, and therefore
source of income. Compared to undergraduate alumni, graduate alumni
miss the opportunity to accrue the private and public benefits asso-
who pursue employment are more likely to report working at a uni- ciated with higher levels of education (Montenegro and Patrinos, 2014;
versity (as teaching or research assistants at North American uni-
Darvas et al., 2017). These findings also support the Sustainable De-
versities, or as lecturers or teaching assistants at African ones) and to velopment Goal (SDG) target 4.B of increasing the availability of
have plans to continue their studies in the near future. More than half of
scholarships to citizens of developing nations, and in particular, African
the 57% of graduate alumni who are working reported that they ap- countries, to increase enrollment in higher education (UN, 2018).
plied to a university program in the 6 months prior to being surveyed,
We did not conduct cost-effectiveness or cost-benefit analyses as
neither were program goals. In its tested form, the Scholars Program is
3
Alumni from Ghana, Kenya, and Uganda have participated in the Program unlikely to be scalable or widely replicated given the generosity of the
through multiple partner institutions, including universities at home and financial and support package—students receive funding for all of their
abroad. Alumni from Lebanon exclusively participated through the American expenses as well as support services. It is also unlikely, however, that
University of Beirut. this level of support would be needed to attain positive impacts on
4
A very small share (3%) earned diplomas, primarily in teaching and nursing. access. This paper showed that the rigorous literature on scholarship

7
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

Table 5
Current country or continent of residence, African Scholars Program alumni (descriptive statistics).
Graduated from a university located in Africa Graduated from a university not located in Africa

Overall
Remained in or returned to Africa 96% 54%
Remained in or returned to home African country 68% 42%
Sample size (N) 68 36

Notes: Results are weighted to adjust for survey non-response. Sample includes Scholars Program alumni who graduated from undergraduate or graduate programs
between June 2015 and July 2016. Universities located in Africa include the African Institute of Mathematical Sciences, Ashesi University, Cape Coast Nursing and
Midwifery Training College, Korle Bu Nursing Training College, Tamale Nursing and Midwifery Training College, and University of Pretoria. Universities located
outside Africa include Arizona State University, Duke University, EARTH University, Michigan State University, Quest University, Savannah College of Art and
Design, St. Olaf College, Trinity College, University of British Columbia, University of California Berkeley, and Westminster College. American University of Beirut
students were excluded from these calculations, as they were not from Africa.

Table 6
Post-Program education and employment outcomes, Scholars Program alumni (descriptive statistics).
Approximately 1.5 years after graduation All Alumni Undergradu- Graduate Female Male
ate Alumni Alumni Alumni Alumni

Enrolled in university 33% 25% 34% 23% 42%


If enrolled:
Enrolled in an honors or master’s degree program 73% 89% 57% 90% 53%
Enrolled in a Ph.D. program 27% 11% 43% 10% 47%
Relying on scholarship support as primary source of funding for graduate studies 57% 50% 66% 68% 53%

Working 61% 70% 57% 73% 53%


If working:
Entrepreneur 7% 4% 9% 5% 10%
Employed by private business 34% 36% 32% 32% 33%
Employed by school or university 25% 20% 30% 23% 32%
Employed by non-profit 16% 20% 13% 18% 16%
Employed in public/government sector 12% 13% 12% 14% 4%
Employed elsewhere 6% 7% 4% 8% 5%

Sample size (N) 124 54 70 75 49

Notes: Results are weighted to adjust for survey non-response. Sample includes Scholars Program alumni who graduated from undergraduate or graduate programs
between June 2015 and July 2016; outcomes are observed approximately 1.5 years after graduation. “Working” includes only alumni who are exclusively working;
alumni who are both enrolled and working are categorized as “enrolled in university.”.

programs is still in its infancy. Testing the impacts of scholarships of- community-based organizations. As the Program evolved, these part-
fering different levels of support will be an important contribution of nerships flourished to enhance the educational opportunities available
future studies. What we can say, however, is that the impacts reported to students and extend the reach of available funding.
here likely reflect the Program’s intentional focus on finding those The experience of the Scholars Program suggests that this ap-
youth that are so disadvantaged or inadequately prepared to find their proach—building partnerships—holds promise for scaling up the use of
way to university that, in the absence of the Scholars Program, would scholarships, as promoted through the SDGs. To succeed in this effort,
not have attended university. Two questions that naturally follow this partnerships may need to consider (1) leveraging actors across sectors
conclusion are (1) how exactly did the Scholars Program achieve this? to implement scholarship programs successfully, (2) developing local
and (2) is the Program’s approach scalable? solutions (programs tailored to the local and, in particular, policy
The Scholars Program was built as a partnership that itself fostered context), and (3) monitoring outcomes through meaningful measures.
new collaborations. The Foundation developed a vision and sought out Below, we consider the experience of the Scholars Program to share
organizations—universities and NGOs—that would help it execute that some thoughts on each of these points.
vision. These organizations needed to provide high-quality education, Although not presented here, qualitative research conducted as part
directly by enrolling and supporting applicants (in the case of university of this study underscored the importance of collaborating with local
partners) or indirectly by helping applicants enroll in universities and actors to recruit the target applicants and correctly interpret the in-
providing other supports throughout their studies (in the case of NGO formation provided in their applications. This is particularly important
partners). These partners also needed to be willing to expand their re- if targeting severely marginalized groups, such as displaced youth; for
cruitment efforts to seek out students that many of them would not example, one Program partner developed relationships with individuals
normally recruit and to expand their linkages with organizations to in refugee camps to identify and assess the special circumstances of
facilitate mentorship, leadership training, service learning, and pro- potential applicants. Partners may also need to broker agreements to
fessional development opportunities, including internships. To do so, share costs—as already done by at least one NGO partner with several
partner universities and NGOs forged new, or strengthened existing, ties universities. Such agreements should consider the danger of new
with local organizations—NGOs, schools, universities, employers, and funding replacing existing funding, as occurred when colleges and

8
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

universities supplanted existing institutional grants with the GMS rates disaggregated by student characteristics such as gender), the labor
scholarship (Amos et al., 2009). Last, partnering with employers to market (through employment rates, for example) and society (through
offer mentoring, shadowing, and internship opportunities can enhance leadership and different forms of civic engagement). Measurement ef-
the educational experience and potentially increase the likelihood of forts may help uncover problems, such as greater university attrition
employment upon graduation (the Program pilot-tested such colla- among students from disadvantaged backgrounds, and weaknesses in
borations and, based on early successful experiences, is currently ex- existing university supports that may help explain those problems, such
panding them). as inadequate or insufficient tutoring or mentoring. Given its scope and
Partnerships may also help foster the development of appropriately commitment to learning activities, the Scholars Program can model
tailored, local solutions to address barriers to access to higher educa- such robust measurement efforts and continue advancing knowledge,
tion. An assessment of local needs, patterns of inequities, and existing particularly about effective policies and factors associated with suc-
policies affecting access can help identify the potential role of scho- cessful outcomes.
larships in a comprehensive strategy to increase equitable participation
in higher education. In particular, understanding other financing op- Funding
tions available to students in the public and private sectors, and po-
tentially coordinating with governments and with donors providing This work was supported by Mastercard Foundation, Toronto,
scholarships, will help identify overlaps (which signal inefficiencies or Canada.
may be needed, given demand), flag gaps in available opportunities,
and prevent unintended consequences. For instance, there is anecdotal Declaration of Competing Interest
evidence that Ugandan students participating in the Scholars Program
at the secondary level (not discussed in this paper) seek to do well None.
enough in the Uganda Advanced Certificate of Education examination
to qualify for the Scholars Program, but not so well that they may earn a Acknowledgments
less advantageous government bursary given to the top students in the
country. Such strategic planning on the part of youth may be difficult to The authors express their sincere appreciation to the many Scholars
prevent, but could translate into government funding going unused. who completed surveys, to the Program partners (colleges and uni-
Designing local solutions—that is, scholarship programming in a col- versities around the world) who participated in interviews and pro-
laborative and contextualized way—may help prevent such in- vided data for analyses, to the staff at Mastercard Foundation who
efficiencies by identifying areas/conditions where the public or the provided insightful feedback on earlier drafts of this paper, to our data
private sector is better suited to provide support. collection partners (Consumer Insights Consult, Research World
Efforts will also need to integrate robust learning and measurement International, and Diane Preciado), and to Mathematica staff who
approaches to monitor progress and improve programs/interventions provided technical assistance, including Ken Fortson, Nick Ingwersen,
over time. This means going beyond basic measures of outputs (such as Jeremy Brecher-Haimson, Seth Morgan, Caroline Lauver, Adele
number of students enrolled or graduated) and efficiency (such as time Rizzuto, and Mary Grider. For more information on the Mastercard
to degree) to measure success in education (for example, graduation Foundation Scholars Program visit: www.mastercardfdn.org/scholars.

Appendix

Institutions included in the study. Table A1 shows the universities in which Scholars enrolled, by whether they are included in the impact and/
or descriptive analysis.
Impact analysis response rates. Table A2 shows the number of Scholars and comparison applicants surveyed and the response rates, overall and
by cohort, for the impact analysis sample. Response rates were higher for applicants in the 2015 cohort, likely due to the shorter time elapsed
between applying to the Scholars Program and receiving the invitation to participate in the survey. Respondents who were excluded from the
analysis were missing values for at least half of the 18 measures that were important predictors of selection for the Scholars Program.
Descriptive analysis response rates. Table A3 shows the number of Scholars Program alumni surveyed and the response rates, overall and by
cohort, for the outcome analysis sample. Response rates were higher for graduate alumni than undergraduate alumni. We excluded from the analysis
respondents who did not complete the first six of the survey’s nine modules. These first six modules covered education, employment, income, living
situation, and reflections on Scholars Program supports.5
Sample exclusions for methodological reasons. Fig. A1 shows the estimated propensity scores for Scholars and comparison applicants. The
green bars show Scholars excluded from the analysis for having a very high probability of selection; these Scholars are not matched to any com-
parison applicants.
Non-response weights. For the descriptive analysis of outcomes data for Program alumni, we constructed non-response weights to account for
observable differences between respondents and non-respondents. Our non-response analysis took into account gender, partner, birth year, gra-
duation year, and type of degree program (undergraduate/diploma or graduate). Using a logit model, we estimated the probability of responding to
the survey – taking into account these characteristics – and then calculated the non-response weight as the inverse of the predicted response
probability using the logit parameter estimates. This approach gives more weight to respondents who have characteristics similar to non-re-
spondents. When estimating means, we calculated the weighted mean applying these non-response weights. In practice, respondents and non-
respondents looked quite similar along these dimensions, so the weighted estimates (which are reported) are similar to the unweighted estimates.

5
The final three modules covered leadership, additional demographics, and contact information.

9
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

Table A1
Scholars’ universities, analytic samples.

(continued on next page)

10
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

Table A1 (continued)

Table A2
Impact analysis sample.
2014 Cohort 2015 Cohort All Total

Scholars Comparison Scholars Comparison Scholars Comparison


Applicants contacted for the survey 217 372 587 610 804 982 1786
Applicants responding to the survey 187 146 547 405 734 551 1285
Response rate 86% 39% 93% 66% 91% 56% 72%
Respondents remaining after excluding those with missing data 176 135 542 382 718 517 1235
Respondents included in matched sample 142 131 495 379 637 510 1147

Notes: Applicants included in the matched sample are those with propensity scores inside the common support (excluded respondents are flagged in green in Fig. A1;
these are Scholars with a probability of selection close to 1).

Table A3
Descriptive analysis sample.
Undergraduate Alumni Graduate Alumni All Tertiary Alumni

Applicants contacted for the survey 127 98 225


Applicants responding to the survey 75 80 155
Response rate 59% 82% 69%
Respondents remaining after excluding those with missing data 54 70 124

11
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

Fig. A1. Propensity scores, treated (Scholars) and untreated (comparison applicants) cases.

References and Mathematics. The Urban Institute and The National Science Foundation,
Washington, DC.
Darvas, Peter, Gao, Shang, Shen, Yijun, Bawany, Bilal, 2017. Sharing Higher Education’s
Al-Samarri, Samer, Zaman, Hassan, 2006. Abolishing school fees in Malawi: the impact on Promise Beyond the Few in Sub-saharan Africa. Directions in Development.
education access and equity. Educ. Econ. 15 (3), 359–375. Washington, DC: World Bank Grouphttps://doi.org/10.1596/978-14648-1050-3.
Amos, Lauren Banks, Windham, Amy M., Brodziak de los Reyes, Iliana, Jones, Wehmah, License: Creative Commons Attribution CC BY 3.0 IGO.
Baran, Virgina, 2009. Delivering on the Promise: An Impact Evaluation of the Gates Deininger, Klaus, 2003. “Does cost of schooling affect enrollment by the poor? Universal
Millennium Scholars Program. Final Report. American Institutes of Research., primary education in Uganda. Econ. Educ. Rev. 22, 291–305.
Washington, DC. Duflo, Esther, Dupas, Pascaline, Kremer, Michael, 2015. Education, HIV, and early fer-
Angrist, Joshua, Bettinger, Eric, Kremer, Michael, 2006. Long-term educational con- tility: experimental evidence from Kenya. Am. Econ. Rev. 105 (9), 2757–2797.
sequences of secondary school vouchers: evidence from administrative records in Filmer, Deon, Schady, Norbert, 2009. School Enrollment, Selection and Test Scores.”
Colombia. Am. Econ. Rev. 96.3, 847–862. Impact Evaluation Series No. 34. Washington. DC: The World Bank Development
Attanasio, Orazio, Fitzsimons, Emla, Gomez, Ana, Gutiérrez, Martha Isabel, Meghir, Research Group – Human Development and Public Services Team July.
Costas, Mesnard, Alice, 2010. Children’s schooling and work in the presence of a Gajigo, Ousman, 2016. Closing the education gender gap: estimating the impact of girls’
conditional cash transfer program in Rural Colombia. Econ. Dev. Cult. Change 58 scholarship program in the Gambia. Educ. Econ. 24.2, 167–188.
(January 2), 181–210. Makundi, Hezron, Huyse, Huib, Develtere, Patrick, Mongula, Benedict, Rutashobya,
Barrera-Osorio, Felipe, Linden, Leigh L., Urquiola, Miguel, 2007. The Effects of User Fee Lettice, 2017. Training abroad and technological capacity building: analyzing the
Reductions on Enrollment: Evidence From a Quasi-experiment. Manuscript. role of Chinese training and scholarship programmes for Tanzanians. Int. J. Educ.
Department of Economics. The University of Texas at Austin. Dev. 57, 11–20.
Behrman, Jere, Parker, Susan, Todd, Petra, 2005. Long-term impacts of the oportunidades Martel, Mirka., 2018. Tracing the spark that lights a flame: a review of methodologies to
conditional cash transfer program on rural youth in Mexico. Gottingen, Germany: measure the outcomes of International scholarship programs. In: Joan, R., Dassin,
Ibero-Am. Inst. Econ. Res. Robin R.Marsh, Mawer, Matt (Eds.), International Scholarships in Higher Education.
Burciul, Barry, Kerr, Kim, 2018. Case study: education in support of social transforma- Palgrave Macmillan, Cham, Switzerland, pp. 281–304.
tion—the Mastercard Foundation scholars program. In: Joan, R., Dassin, Robin Mawer, Matt., 2014. A Study of Research Methodology Used in Evaluations of
R.Marsh, Mawer, Matt (Eds.), International Scholarships in Higher Education. International Scholarship Schemes for Higher Education. Commonwealth Scholarship
Palgrave Macmillan, Cham, Switzerland, pp. 329–343. Commission in the UK., London.
Breeze, Victoria, Moore, Nathan, 2017. China Has Overtaken the US and UK As the Top Mawer, Matt., 2017. Approaches to analyzing the outcomes of international scholarship
Destination for Anglophone African Students. The Conversation. programs for higher education. J. Stud. Int. Educ. 21 (3), 230–245.
Canning, David, Raja, Sangeeta, Yazbeck, Abdo S., 2015. Africa’s Demographic Transition: Meng, Xin, Ryan, Jim, 2010. Does a food for education program affect school outcomes?
Dividend or Disaster? Overview Booklet. Africa Development Forum Series. World The Bangladesh case. J. Popul. Econ. 23 (March 2), 415–447.
Bank, Washington, DC. https://doi.org/10.1596/978-1-4648-0489-2. License: Montenegro, Claudio E., Patrinos, Harry Anthony, 2014. Comparable Estimates of
Creative Commons Attribution CC BY 3.0 IGO. Returns to Schooling Around the World. Policy Research Working Paper 7020.
Clewell, Beatriz Chu, de Cohen, Clemencia Cosentino, Tsui, Lisa, Deterding, Nicole, 2006. Education Global Practice Group. World Bank, Washington, DC.
Revitalizing the Nation’s Talent Pool in STEM—Science, Technology, Engineering, Perna, Laura W., Orosz, Kata, Gopaul, Bryan, Jumakulov, Zakir, Ashirbekov, Adil,

12
C. Cosentino, et al. International Journal of Educational Development 71 (2019) 102089

Kishkentayeva, Marina, 2014. Promoting human capital development: a typology of United Nations Educational, Scientific, and Cultural Organization Institute for Statistics
international scholarship programs in higher education. Educ. Res. 43 (2), 63–73. (UIS). “2017 estimates.” UIS.Stat database, n.d. Available at http://data.uis.unesco.
United Nations. “Population.” n.d. Available at http://www.un.org/en/sections/issues- org. Accessed November 18, 2018.
depth/population/. Accessed November 18, 2018. United Nations, 2018. The Sustainable Development Goals Report 2018. United Nations,
United Nations Educational, Scientific, and Cultural Organization Institute for Statistics Division for Sustainable Development Goals, New York. https://
(UIS), 2010. Trends in Tertiary Education: Sub-Saharan Africa. UIS Fact Sheet sustainabledevelopment.un.org/sdg4.
No.10. .

13

You might also like