Social Impact Bond in Emerging Market: Funding Chinese Migrant Workers’ Children Digital Education
David Ng1 Professor of Finance, College of Business
Cornell University Ithaca, NY
Bill Tsang Research Associate, University of Hong Kong, and Research Director, Hong Kong Youth Foundation
Hong Kong, China
1 David Ng acknowledges generous financial support from Roberto Cañizares Fund at the Emerging Market Institute at Cornell’s Johnson School of Graduate Management, and also from Cornell’s Lehman Fund for Scholarly Exchange with China.
Social Impact Bond in Emerging Market: Funding Chinese Migrant Workers’ Children Digital Education
Abstract:
Social impact bonds carry the promise of bringing private capital to achieve social goals in different countries. Social impact bond is a contractual agreement where government (or international organization) promises to pay when specific social outcome benchmarks are achieved. Private investors would invest to provide current funding for the social intervention, and get compensated with repayment and returns once the benchmarks are achieved. This paper examines the feasibility of a program that uses digital media to supplement and facilitate the education of migrants’ children in China. We conducted a pilot study involving a migrant workers’ children school in Beijing and examined the test score improvements for hundreds of elementary school students. We identified and classified free educational videos on the web, organized them at grade-level, and distributed the videos to migrant workers' children in handheld tablets. We monitored the test score improvements to the students before and after they received the tablets, and compared such improvements to students who did not have the tablets. Overall, we find that students benefited substantially from the math content in the tablets. We propose a social impact bond structure on financing an expansion of the program.
I. Introduction
Traditional approaches to financing development projects in emerging countries rely on
aid and loans from industrial countries, the World Bank, and private foundations.
However, such channels are not sufficient and face great challenges. Industrial countries’
governments have less interest in providing foreign aid in the face of severe budget
constraints and debts. As Karolyi (2014) points out, the line is blurred between emerging
countries and industrial countries. Countries like China are getting richer, and are no
longer qualified for foreign aid on a governmental level. However, specific segments in
China would still desperately need capital for beneficial developmental projects.
With government budgets tightening worldwide, there is a need to channel limited
government and international organization funding towards high-quality developmental
projects. A new type of investment vehicle, social impact bond, carries the promise of
bringing private capital to achieve important social objectives. Social impact bond is a
contractual agreement where government (or international organization) promises to pay
when specific social outcome benchmarks are achieved. Private investors would invest
to provide current funding for the social intervention, and get compensated with
repayment and returns once the benchmarks are achieved.
Social impact bonds carry the promise of bringing private capital to achieve
developmental goals in emerging market countries. The social impact bond brings
positive benefit for private investors, governments and the social enterprises. Private
investors in social impact bonds potentially earn returns and provide screening and
monitoring. Governments would be able to reduce risk and pay only when a successful
outcome happens. Social enterprises would receive up front capital to provide social
services.
Migrant workers’ children’s education is a challenging issue in China. As we explain in
detail later, as a result of regional economic discrepancy there are millions of migrant
workers in Chinese coastal cities. Migrant workers’ children do not have rights to receive
education in regular city elementary and junior public schools, which are free and
guaranteed for all other school-age children.
By law, none of the migrant workers’ children are allowed to take the local exam to enter
university. In fact, few of them even get to go to junior high school. These children could
return to study in their home provinces but they would be separate from their working
parents during junior high school or high school years. However, after receiving low-
quality elementary school education, the children could not catch up. Most of those who
return either drop out of school or return to the city within a year or two. (Koo, Ming, &
Tsang, 2014; Ming, 2014)
This paper proposes a potential social impact bond project to provide digital media
education for migrant workers’ children in China. We explore the feasibility of using
digital media to supplement and facilitate the education of these migrants’ children.
The rest of the paper is divided into four sections. Section II introduces the financial
structure of a social impact bond and provides an example of social impact bond. Section
III explains the background of migrant workers’ children education in China. Section IV
describes our experiment. Section V proposes a social impact bond structure for such a
project. Section VI concludes.
II. Background on Social Impact Bond
A social impact bond involves a government, service providers and private investor. Take
the following illustration as an example.
Figure 1: Structure of a Social Impact Bond
Private investors invest a certain amount of capital into a social impact bond. This sum of
money serves as the working capital for a service provider to serve a targeted population.
After a period of time, once certain performance benchmark is reached, a payer, e.g. the
government, will realize the cost saving and provide performance-based payment to the
social impact bond investors. The private investors then obtain repayment of their capital
plus some returns. If the performance benchmark is not reached, then there will be no
payment, and the social impact bond investors will suffer from losses.
Recent initiatives include, for example, financing early childhood education, improving infant mortality rate, or reducing recidivism rate. Harvard Kennedy School has a Social Impact Bond Technical Assistance Lab, led by Professor Jeffrey Liebman. A startup organization, Instiglio, specializes in bringing social impact bonds to developing countries.
The main risk facing social impact bond investors is that they will not get paid for the
principal or interest, if there is insufficient reduction to special education. From the
investor’s point of view, a social impact bond could also provide diversification potential
since the outcome is based on the success of a particular social outcome, which has little
correlation with returns in other asset classes such as equities or bonds. If social impact
bonds should become a big market, this could potentially attract capital investment from
institutional investors. There are variations to the social impact bond I describe above.
For example, MacLean (2013) proposes a structure in the form of a futures contract.
The first social impact bond on early childhood education in the U.S. was issued in 2013
in Utah. (Popper 2015). In the first year of implementation, 595 low-income children got
to receive pre-school education, out of which 110 had been expected to need special
education when they reached school age if they did not attend preschool. Special
education is catered towards children with special needs and requires individual
educational plans and teaching resources, and is an expensive item for Utah government.
A high quality pre-school education for the children may speed up learning capability of
the children and reduce the subsequent need for special education. Such reduction can
potentially save the state government a substantial amount of money.
Utah is willing to pay a lump sum, but only if special education need is reduced down the
road. If the special education rate is reduced to the targeted level by a certain date, this
will trigger a payment from the Utah government, which will cover the principal
repayment and provide a return for the private investors. For each pre-school child that
took an initial test and was deemed likely to need special education, the state would pay
the social impact bond investor $2500 each year that the child managed to avoid special
education.
This promise of future payout becomes an illiquid asset against which the social impact
bond is issued. In this initial round, Goldman Sachs lends $4.6 million and serves as the
underwriter for the issue, while Pritzker Family Foundation invests $2.4 million, . This
money helps pay for the expansion of preschool education to more children. According to
Popper (2015), the Utah social impact bond project had a successful outcome. Out of the
110 students considered likely to need special education at the end of preschool, only 1
student eventually needed special education. As a result of this success, Goldman Sachs
was paid the principal plus a $1 million returns.
A social impact bond can potentially provide positive outcomes to all parties involved.
Utah government can achieve cost saving at no risk; it only pays once the saving from
special education is realized. The school receives immediate working capital to scale up
effective pre-school education programs. Pre-school aged children get high quality pre-
school education from the school to help them stay out of subsequent special education.
Private investors get to receive potential returns in their investments. The incentives are
aligned to achieve good social outcomes. Private investors have incentives to invest in
effective schools and provide good monitoring to the schools. The school also has the
incentive to use the money effectively to achieve a better outcome, especially when
compared to the alternative case when there is no string attached with a donation. A
vibrant social impact bond market can stimulate careful research, quantification and
implementation on a particular social issue.
While the Utah social impact bond is considered a success, not all social impact bonds
are quite as successful. A social impact bond that was designed to reduce recidivism rates
in New York Rikers’ Islands prison had a high profile launch in 2013. (Ohlson and
Philips 2014) Despite a lot of good will, work and investment by the multiple parties
involved, the social impact bond failed after a couple years due to poor execution, low
participation, and poor documentation of participation vs non-participation in the
program. (New York Times 2015)
Only a few social impact bonds have been created in developing countries. There is
active interest from foundations (see Social Finance (2013)) in promoting such efforts.
This project is exploratory and involves research associated with a potential project that
may be feasible for a social impact bond. In particular, I examine a specific application
on an on-line education program for Chinese migrant workers’ children.
III. China migrant workers’ children education
China is known as a major manufacturing hub to the world. Most of the Chinese goods
are produced by factory workers that migrate from the countryside to cities like Beijing,
Shanghai or Guangzhou. Such migrant workers work at low wages and poor working
conditions and live a Spartan existence. Because such migrant workers have no local
resident status (hukou), they enjoy no social security, health care, and unemployment
benefits from the city, even though they do pay taxes. What’s even more appalling is that
their children (over half a million of them in Beijing alone, and 30 millions around the
country) have very little educational opportunities.
Migrant workers’ children do not get education in regular elementary and junior public
schools in the city, which are free and guaranteed for all other children at the elementary
and junior high school level. If the children stay in the city, they only get to go to some
make-shift migrant workers’ schools during elementary school years. These private
schools operate without proper licenses in the cities. Often times, the schools will be
closed down by the city government for some arbitrary reasons. For instance, in August
2011, “thousands of migrant workers' children in Beijing have been left with no school to
attend after officials abruptly closed their schools. State media said some 14,000 children
in three districts of the Chinese capital had been hit.” (British Broadcasting Company,
2011)
Under such constant threats, many migrant schools would skimp on facilities and
teachers’ salaries and operate on a minimum subsistence level. In most schools, six
primary school grades are combined into just two grades to save on resources. Some
schools turn off the lights in the classrooms to save electricity. Teachers are paid poorly,
have extremely high turnover and barely teach basic materials. Schools would pack up
and move to another make-shift sites once they are closed down by the authorities. By
law, none of the migrant workers’ children are allowed to take the local exam to enter
university. In fact, few of them even get to go to junior high school. Migrant workers’
kids could leave their parents and return to the countryside to study there on their own
during junior high school or high school years. But as these kids have low-quality
elementary school education, they have problems catching up. Previous research (Koo,
Ming, & Tsang, 2014; Ming, 2014) shows that most of those who return either drop out
of school or return to the city within a year or two.
Under such situations, the students do not learn while they grow up; they merely grow
up. A field worker says that in the first year of junior high, migrant workers’ children
would feel the same as other kids and aspire to be doctors and lawyers. But when such
children advance in years, they realize they have not learned anything and have no hope
of getting more education, and begin to settle to become a peasant or a factory worker.
IV. Pilot Project on Proposed Intervention: Digital Media To Improve Learning
A. Setup
This pilot project evaluates using digital media to improve the educational quality of the
migrant workers' children in China. The two main researchers involved are David Ng
from Cornell Dyson School and Bill Tsang from Hong Kong Youth Foundation (HKYF)
and Hong Kong University. Serving as the main coordinator on the ground was Bill
Tsang, a Harvard-trained psychologist focusing on Chinese migrant workers’ children.
He and his colleagues coordinated research assistants, graduate students and volunteers
on the ground in Beijing.
To conduct a pilot study, we identified and classified free educational videos on the web,
organize them, and distributed the videos to migrant workers' children into a handheld
device. These videos were used in classroom or home settings. We then used a pilot
program to compute test score improvements to the children before and after the videos.
Our goal was to assess whether an education via digital media can help to equalize the
educational outcomes of migrant workers’ children and those of regular school-aged
children. After testing a few different kinds of tablets, we decided to use a no-frill hand-
held 7-inch tablet produced in China. The tablet works on an Android platform. See
figure 4. The average cost of each tablet was about $60. We bought 56 tablets in total for
the first stage.
We recruited a computer programmer, who helped us to create the programming
interface. Because the tablets were inexpensive, the failure rates were relatively high. Our
programmers also took care of some of the repairs.
We downloaded educational materials from the web. There are some free educational
materials on the web for children. A case in point is Khan Academy,
www.khanacademy.org, which is a website with over 2800 education videos for children
to learn. These bite-sized, short 10-minute videos cover a wide range of materials in
mathematics, physics, chemistry, biology, economics, history and various other
disciplines. Mathematics, for instance, starts from basic addition and goes all the way up
to college-level materials. Since much of Khan Academy’s content either lacked Chinese
translation or just had Chinese subtitles, we downloaded reading materials on science and
languages from various other Chinese websites.
We found different educational videos for children of this age, based on math curriculum
on their grade-level. The topics included: addition and subtraction, multiplication and
division, factors and multiples, negative numbers and absolute values, decimals and
percent, fractions, and arithmetic properties. Each topic involved about 20+ videos, many
of which are from Khan Academy. Figure 4 shows a list of videos for decimals and
percent.
We conducted the project in a school called Hua Ao School, a school in the outskirt of
Beijing. See Figures 2 and 3. It is one of a few relatively established migrant schools in
Beijing. We conducted three pilot tests over the course of two years. Pilot 1 started from
March to June in 2013 and involved grade 6 students only. Pilot 2 started from October to
December in 2013, and involved students from grade 4 to grade 6 students. Pilot 3 started
in 2014 from March to June, and involved grade 4, 5, and 6 students. Each Pilot lasted
about 10 weeks.
Students voluntarily signed up to be selected into receiving the tablets during the pilot
projects. The teachers were responsible for distributing the tablets. Each tablet was
shared by two students. Each student got to keep the tablet for a few days at a time to
watch the video and read the materials on the tablet.
B. Results on the Pilot Project
There are three pilots in total. In Pilot 1, we provided only Math video content and tested
the students on Math only, and only for 6th grade students. In Pilot 2, we expanded and
tested students on Math for 4th, 5th and 6th grade students. In Pilot 3, we conducted tests
on Math, English and Chinese, and tested students on 4th, 5th and 6th grades.
In pilots 1 and 2, we conducted a test before the distribution of the tablet, and then
conducted another test with the same content two months after the distribution of the
tablet. Over these two months, students were still taking regular classes, so they should
have better math skills afterwards. Hence, we expected the control group of student (i.e.
students in the same class without the help of the tablet) would improve test scores.
What’s more interesting is to see whether the treatment group of students (i.e. those
students with the tablet and math videos) would have even higher test scores than the
control group of students. Pilot 3 is similar to pilots 1 and 2, except that due to the
scheduling issue, we distributed the tablets right before the midterm, but after the first
two tests in the semester. We compared the final scores with the scores the students
received for the average of the first two tests.
Table 1 documents the pooled results for all three pilots and for all students in all grades.
Overall, 179 students were in the test group and had access to tablets, while 335 students
were in the control group. For the test group of students, students received a baseline
score of 62.98, which is very similar to the control group of students who had a baseline
score of 62.47. After receiving the tablets, the test group of students saw an improvement
in score of 3.28, as compared to the control group of students who saw an improvement
in score of only 0.33. The difference in improvement is statistically significant at a 5%
significance level with a t-statistics of 1.96.
The bottom panel shows the improvement by grades. The main improvement is for grade
6 students who are by far the largest group of students. For grade 4 and grade 5 students,
there were no improvements.
To see the results for the three pilot results separately, we turn to Table 2. Table 2 Panel
A shows the results for Pilot 1. In pilot 1, there are 105 6th grade students in total, of
whom 61 students would have access to the tablet and its math educational materials. As
expected the test scores have improved by 3.44 for the control group of students who had
no tablet. Those students with the help of the tablets saw their test scores improved
substantially more by about 9.74. The difference of the improvement scores is 6.29 (over
10% of the score). Because of the low number of students (only 61), the p-value is only
0.15 and is not statistically significant at a 10% level. However, the relatively large
improvement of test score suggests that there is good potential for the tablet.
Table 2 Panel B shows the results for Pilot 2. In pilot 2, we would like to see if there is
any difference for students in different grades. We distributed the same sets of tablets to
students in three different grades (grade 4, 5, and 6) and increased math contents to
grades 4 and 5. Table 2 Panel B basically shows the grades are the same among the
intervention and the control groups for all three grades.
Table 3 reports the results for pilot 3. The results shows that for grade 6 students, the
intervention group with the tablet shows large improvement in grades after the
intervention. The test score improves by 10.6 points (over 16 percent of the score), and is
statistically significant at 3.22. For grade 4 and 5 students, there is no improvement.
We interpret the results to be that grade 6 students may be more mature to self-learn,
while younger children may need more handholding in a classroom setting.
Overall, the significant improvement in math test score for grade 6 students makes us
believe that there is more room for digital learning to improve education outcome among
Chinese migrant workers’ children.
Table 3 documents the effects for English and Chinese videos, which are part of a test for
Pilot 3. We find no significant improvement in test scores for the children. It could be
due to the subject matter; English and Chinese may be less well-defined than the math
curriculum. It could also be due to the fact that we cannot find close match between video
content. In any case, we do not find much effect for these language subjects.
In sum, we find that the usage of digital media has potential to improve the educational
outcomes of migrant workers’ children in China. We hope to expand the number of
students, classrooms and schools but we are constrained by the tablets and additional
personnel costs.
C. Discussion on Implementation
As we noted earlier, success of a social impact bond project depends on successful
implementation. During the course of the implementation of the pilot programs, there
were many challenges that we encounter. Some parents were concerned that children
would study less when they got hold of a tablet computer. In fact, the concerns were at
least partially justified. At the start of the pilot, the tablet computer only contained
education videos. Students soon figured out how to take pictures and downloaded
computer games using the tablet, and spent a lot of time on that. Some even did it in
class, which was eventually stopped by the teachers.
Subsequently, we asked the programmer to create a restriction on the tablet to exclude
picture taking and downloading of computer games. This worked for several weeks, after
which a student figured out how to break the restriction by repeatedly pressing a certain
key. He then shared this with the entire class. Eventually, we successfully created a
secured restriction and students no longer could break the restriction.
Another issue is that some of the students were disruptive by using the tablet during
classes. Also, hardware problems occurred in quite a few machines, and repairs were
quite frequent. This was partially exasperated by the fact that some students did not really
take good care of the tablet, which led to hardware problems in some machines.
Eventually, we asked students to put in a nominal fee (US$16, equivalent of 100 RMB as
deposit that would be refunded to them if the machine was returned in good order. This
made them feel more responsible for the machine and reduced the repair rates.
We encountered other problems with the test evaluations as well. As migrant workers’
children came from unstable families. Some children might have to drop out or transfer
from the current school when their parents could not find work. This led to a problem in
our experimental design, as some students were not able to complete the post-test
evaluation before the left school..
VI. A Proposed Social Impact Bond
Based on the pilot study we run, we propose a social impact bond that has the following
structure.
The total working capital we propose is $200,000 up front. This would allow the pilot
project to expand to multiple schools. Private investors would put up the upfront capital.
To compensate private investors, a foundation would provide a $230,000 pool of money
to be distributed in case of successful performance over two years time.
The performance standard is based on math test scores by sixth grade graduation. We
propose that if more than 70% of students see an improvement in test scores, then full
payment of $230,000 would be triggered. If between 40% and 70% of students see an
improvement in test scores, then the investor will be paid $210,000. If 25% to 40% see an
improvement of test scores, then the investor will be paid $190,000. If fewer than 25% of
students see an improvement in test scores, then the investors will be paid $100,000.
This payment schedule means that much of the risk is born by the government who is
responsible for half the payment. But at least some of the risk is borne by the private
investor.
The choice of using test score improvement deserves some discussions. We do not
believe that students’ improvements in learning capability can be completely measured
by test scores. At a minimum, test score improvement provides one measurable
benchmark based upon which some cognitive improvements can be measured. Also, tests
play important roles in Asian society in terms of students’ abilities to further their
educations at least in their home towns.
One feature of social impact bond is to identify a long-term payer who realizes the
benefits from the proposed intervention and promises to pay. Typically, for social impact
bond in the U.S. or U.K., the government realizes the benefit from the intervention and
promises to pay. Migrant workers’ children are not recognized by the cities where their
parents work. They are also not physically present in the home provinces where their
parents came from, and hence would not get support from these provinces. In the current
social impact bond involving migrant workers, the only, very remote possibility is the
Chinese central government. In the short term, it may be more realistic to locate a
foundation to promise to pay for the larger pilot programs.
The advantage of this project for the social impact bond is that the implementation of the
intervention has been conducted before in a pilot. There is a working partner organization
and a partner school. This specific intervention has been tested in the field so
implementation would not be completely new. The main thing that still needs to be done
is the social impact bond structure.
VI. Conclusions and extensions
This paper discusses the applicability of social impact bond in a particular setting in
China. But social impact bonds can have much broader appeal. Social impact bond can
potentially attract substantial private capital to achieve important social objectives in
emerging markets. It can help to allocate resources effectively and align incentives well.
As Jeffrey Sachs (2014) states, “the issue is not "yes" or "no" to aid. Aid is needed, and
can be highly successful. The issue is how to deliver high-quality aid to the world's
poorest and most vulnerable people.” A social impact bond targeted towards specific
need may provide a better structure for providing aid to developing countries.
As these types of vehicles are just emerging in the investment community, there is little
research on the topic. Yet, this topic is important and could be of interest for many
parties: finance researchers, development and empirical economists, social entrepreneurs,
governments, institutional investors, potential donors and students. With its deep
expertise in various facets of emerging markets, Cornell has a comparative advantage in
this line of research.
The social impact bond structure can be applied in other settings as well. Think of a
potential application of AIDS reduction for a South African community. The prevalence
of AIDS has devastated many parts of the country and is expensive to treat. People can
contract AIDS through either unsafe sex or blood contact. Left untreated, AIDS is
contagious and can spread rapidly. Diseases such as these also affects companies’ bottom
lines when they lose workers, suppliers, and clients to the disease. This forces companies
to have to invest more in training new workers, creates more turnover, and can decrease
the productivity of companies.
The best way to deal with AIDS is through effective prevention and early treatment. If an
NGO can organize, educate, and arrange treatment for a community, it could help contain
the spread of the disease. The groundwork for these NGOs are in place currently, but they
do not have funding to scale up their operations.
Social impact bonds can be useful in this regard. Companies have incentives to pay for a
reduction in AIDS rate by a certain date in the community they operate. Additionally,
outside donors or foundations may also promise grant payment. The Non-governmental
Organizations (NGOs) can use the promised payment as an asset to issue a social impact
bond. Private investors can then invest in the social impact bonds to finance the
operations of NGOs.
There are various challenges in such a bond. For instance, the legal environment and
contract enforcement for bond issues in South Africa is not as comprehensive as in
developed countries such as the United States. Furthermore, high inflation may eat away
any returns investors may have. Not many private investors are interested in bonds with
such issues, and potential investors may encounter higher risks in certain countries.
While there is risk, there are also a lot of promises. Social impact bonds can potentially
be issued in many different areas. Similar investment vehicles can be structured towards
other social issues like improving infant mortality or building more schools in
communities in developing countries.
Figure 2. Left Top picture. Students receive the tablets. Left Right picture. Students try out the new tablets. Left Bottom picture. Bill Tsang and David Ng in front of Hua Ao School. Right Bottom picture. Vice principal, David Ng, and guidance counselor of Hua Ao School.
’’
Figure 3. Different parts of Hua Ao school
Figure 4. A sample of math videos we documented.
Table 1: Pooled test results for all three pilots
Group size
Final score
Baseline score
Final score minus baseline score
All
control group(tablet=0) 335 62.81 62.47 0.33 test group(tablet=1) 179 66.26 62.98 3.28 improvement 3.45 0.50 2.95 improvement % 0.05 0.01 improvement % diff 0.02 t (0.23) (-1.47) (1.96) p-value 0.82 0.14 0.05
Grade 4
control group(tablet=0) 99 67.73 71.79 -4.05 test group(tablet=1) 34 69.00 74.95 -5.95 improvement 1.27 3.17 -1.90 improvement % 0.02 0.04 improvement % diff -0.06 t (0.24) (0.74) (-0.68) p-value 0.81 0.46 0.50
Grade 5
control group(tablet=0) 83 55.75 58.01 -1.84 test group(tablet=1) 37 65.75 67.98 -2.23 improvement 9.99 9.97 -0.39 improvement % 0.18 0.17 improvement % diff 0.01 t (1.72) (1.99) (-0.13) p-value (0.09) 0.05 0.90
Grade 6
control group(tablet=0) 173 63.32 59.24 4.08 test group(tablet=1) 108 65.56 57.39 8.17 improvement 2.23 -1.85 4.09 improvement % 0.04 -0.03 improvement % diff 0.04 t (0.76) (-0.70) (1.99) p-value 0.45 0.49 0.05
Table 2 Panel A: Pilot 1 Math test results
Grade 6
Group size Final Baseline Final minus
baseline control group(tablet=0) 44 58.61 55.17 3.44 test group(tablet=1) 61 69.30 59.56 9.74 improvement
10.69 4.40 6.29
improvement %
0.18 0.08 improvement % diff
0.10
t
(1.46) p-value 0.15
Table 2 Panel B: Pilot 2 Math test results
Grade 4
Group size Final Baseline Final minus
baseline control group(tablet=0) 49 64.95 64.53 0.42 test group(tablet=1) 16 59.28 59.70 -0.42 improvement
-5.67 -4.83 -0.83
improvement %
-0.09 -0.07 improvement % diff
-0.01
t
(-0.17) p-value
0.87
Grade 5
control group(tablet=0) 46 33.07 35.31 -2.24 test group(tablet=1) 13 32.17 36.60 -4.43 improvement
-0.90 1.29 -2.19
improvement %
-0.03 0.04 improvement % diff
-0.05
t
(-0.38) p-value
0.71
Grade 6
control group(tablet=0) 56 46.70 42.75 3.95 test group(tablet=1) 28 41.88 41.56 0.32 improvement
-4.82 -1.19 -3.63
improvement %
-0.10 -0.03 improvement % diff
-0.08
t
(-1.17) p-value
0.24
All
control group(tablet=0) 151 48.47 47.55 0.92 test group(tablet=1) 57 44.55 45.52 -0.97 improvement
-3.92 -2.03 -1.89
improvement %
-0.08 -0.04 improvement % diff
-0.04
t
(-0.81) p-value 0.42
Table 2 Panel C: Pilot 3 Math test results
Grade 4
Group size Final Baseline Final minus
baseline control group(tablet=0) 50 71.91 78.90 -6.98 test group(tablet=1) 18 77.64 88.51 -10.88 improvement 5.72 9.62 -3.89 improvement % 0.08 0.12 improvement % diff -0.02 t (-0.97) p-value 0.34
Grade 5
group size final minus
baseline control group(tablet=0) 37 85.81 87.02 -1.21 test group(tablet=1) 24 83.94 84.98 -1.04 improvement -1.87 -2.04 0.17 improvement % -0.02 -0.02 improvement % diff 0.00 t (0.08) p-value 0.94
Grade 6
group size final minus
baseline control group(tablet=0) 73 82.10 77.15 4.95 test group(tablet=1) 19 91.39 75.50 15.89 improvement 9.29 -1.65 10.94 improvement % 0.11 -0.02 improvement % diff 0.16 t (3.22) p-value 0.00
All
group size final minus
baseline control group(tablet=0) 160 79.62 79.94 -0.32 test group(tablet=1) 61 84.28 83.20 1.09 improvement 4.67 3.26 1.41 improvement % 0.06 0.04 improvement % diff -0.04 t (0.81) p-value 0.42
Table 3: Pilot 3 English and Chinese test results Panel A: Pilot 3 English test
Grade 4
Group size Final Baseline Final minus
Baseline control group(tablet=0) 50 65.81 58.76 7.05 test group(tablet=1) 18 76.72 68.06 8.67 improvement
10.92 9.30 1.62
improvement %
0.17 0.16 improvement % diff
-0.01
t
(0.42) p-value
0.68
Grade 5
control
group(tablet=0) 37 78.87 62.74 16.13 test group(tablet=1) 24 81.44 66.33 15.11 improvement
2.57 3.60 -1.02
improvement %
0.03 0.06 improvement % diff
-0.04
t
(-0.31) p-value
0.76
Grade 6
control
group(tablet=0) 73 73.38 60.36 13.02 test group(tablet=1) 19 85.28 71.82 13.46 improvement
11.90 11.46 0.44
improvement %
0.16 0.19 improvement % diff
-0.04
t
(0.13) p-value
0.90
All
control
group(tablet=0) 160 72.14 60.40 11.74 test group(tablet=1) 61 81.18 68.50 12.68 improvement
9.04 8.10 0.94
improvement %
0.13 0.13 improvement % diff
-0.02
t
(0.45) p-value 0.66
Panel B: Chinese test
Grade 4
Group size Final Baseline
Final minus
Baseline control group(tablet=0) 50 77.07 75.56 -0.13 test group(tablet=1) 18 83.56 88.36 -4.81 improvement
6.49 12.80 -4.68
improvement %
0.08 0.17 improvement % diff
-0.05
t
(-2.09) p-value
0.04
Grade 5
control
group(tablet=0) 37 85.82 86.29 -0.13 test group(tablet=1) 24 88.63 89.14 -0.51 improvement
2.80 2.84 -0.38
improvement %
0.03 0.03 improvement % diff
0.00
t
(-0.03) p-value
0.98
Grade 6
control
group(tablet=0) 73 89.25 73.91 14.15 test group(tablet=1) 19 92.00 76.03 13.53 improvement
2.75 2.12 -0.62
improvement %
0.03 0.03 improvement % diff
0.01
t
(-0.17) p-value
0.87
All
control
group(tablet=0) 160 84.40 77.23 5.86 test group(tablet=1) 61 88.12 84.82 2.41 improvement
3.72 7.59 -3.45
improvement %
0.04 0.10 improvement % diff
-0.04
t
(-0.96) p-value 0.34
Bibliography British Broadcasting Company, 2011, Migrant Schools Closed in Chinese Capital, http://www.bbc.com/news/world-asia-pacific-14556906 Karolyi, Andrew, 2015, Cracking the emerging market enigma, Oxford University Press. Koo, Anita, Ming, Holly, and Tsang, Bill, 2014, The doubly disadvantaged: How return migrant students fail to access and deploy capitals for academic success in rural schools. Sociology, 48(4), 795-811. Ming, Holly, 2014, The Education of Migrant Children and China’s Future, Routledge, New York, NY. MacLean, Guillermo, 2013, Understanding DreamFutures Contracts, Securities and Markets. Olson, John and Andrea Phillips, 2014, Rikers Island: The First Social Impact Bond in the United States, Federal Reserve Bank of San Francisco Community Development Investment Review. Popper, Nathaniel, 2015, For Goldman, Success in Social Impact Bond that Aids Schoolchildren, New York Times. Social Finance, 2013, A New Tool for Scaling Impact: How Social Impact Bonds Can Mobilize Private Capital to Advance Social Good, working paper. Sachs, Jeffrey, 2014, The case for aid, Foreign Policy, January 21 issue. New York Times, 2015b, Wall Street Money Meets Social Policy at Rikers Island, July 28.