Can One Laptop per Child Reduce the Digital Divide and Educational Gap? Evidence from a Randomized Experiment in Migrant Schools in Beijing Di Mo, Johan Swinnen, Linxiu Zhang, Hongmei Yi, Qinghe Qu, Matthew Boswell, Scott Rozelle Abstract One Laptop per Child (OLPC) is one of the high profile initiatives to try to narrow the inequality of access to ICT (digital divide). However, despite the fact that OLPC currently has distributed more than two million laptops in more than 40 countries, there is little empirical evidence that is available to help us understand the impacts of the program. The goal of our study is to assess the effectiveness of OLPC in narrowing the digital divide between poor and rich children in China and in increasing the human capital of disadvantaged children. In order to do so, we conducted a randomized experiment involving 300 third-grade students in 13 migrant schools in Beijing. Our results show that, the OLPC program improved student computer skill scales by 0.33 standard deviations and standardized math scores by 0.17 standard deviations after 6 months of intervention. Less-skilled students improved more in computer skills after the program. Moreover, the OLPC program also significantly increased student learning activity using computer software and decreased the time students spent watching TV. Students’ self-esteem also improved with the program. Working Paper 233 March 2012 reapchina.org/reap.stanford.edu
41
Embed
Is One Laptop per Child Reducing Digital Divide and ...fsi.stanford.edu/sites/default/files/olpc_paper_March_31_2012_Web.pdf · Can One Laptop per Child Reduce the Digital Divide
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Can One Laptop per Child Reduce the Digital Divide and
Educational Gap? Evidence from a Randomized Experiment in
Migrant Schools in Beijing
Di Mo, Johan Swinnen, Linxiu Zhang, Hongmei Yi, Qinghe Qu, Matthew Boswell,
Scott Rozelle
Abstract
One Laptop per Child (OLPC) is one of the high profile initiatives to try to narrow the
inequality of access to ICT (digital divide). However, despite the fact that OLPC
currently has distributed more than two million laptops in more than 40 countries,
there is little empirical evidence that is available to help us understand the impacts
of the program. The goal of our study is to assess the effectiveness of OLPC in
narrowing the digital divide between poor and rich children in China and in
increasing the human capital of disadvantaged children. In order to do so, we
conducted a randomized experiment involving 300 third-grade students in 13 migrant
schools in Beijing. Our results show that, the OLPC program improved student
computer skill scales by 0.33 standard deviations and standardized math scores by
0.17 standard deviations after 6 months of intervention. Less-skilled students
improved more in computer skills after the program. Moreover, the OLPC program
also significantly increased student learning activity using computer software and
decreased the time students spent watching TV. Students’ self-esteem also improved
with the program.
Working Paper 233
March 2012
reapchina.org/reap.stanford.edu
January 2012
Can One Laptop per Child Reduce the Digital Divide and Educational Gap?
Evidence from a Randomized Experiment in Migrant Schools in Beijing
Mo Di and Johan Swinnen
LICOS Centre for Institutions and Economic Performance, KU Leuven
Linxiu Zhang, Hongmei Yi and Qinghe Qu
CCAP, IGSNRR, Chinese Academy of Sciences
Matthew Boswell and Scott Rozelle
Stanford University
Acknowledgements:
We wish to acknowledge the support of Quanta Computing, Tianhua Shidai, TAG
Family Foundation, Bowei Lee and Family and Mary Ann Millias St. Peters for
support for this project. The efforts of many others also made it possible, including,
Shufen Chen, Lei Jiang, Tianxi Lin and Xiao Hu.
Can One Laptop per Child Reduce the Digital Divide and Educational Gap?
Evidence from a Randomized Experiment in Migrant Schools in Beijing
Access to and facility with information and communication technology (ICT)
in the 21st century is increasingly important for individuals. ICT adoption is costly,
but there are direct benefits to the easier and faster access to information that it
affords. This remains true even if the degree of adoption is less than ideal (Jensen,
2007). For instance, only individuals that have accurate information about prices can
engage in efficient trading. Individuals with good information can often better utilize
social services, such as education and health. Good information is more accessible
when an individual is more facile with ICT (Cilan et al., 2009). ICT has also become
an important contributor to growth in productivity and a nation’s overall economy
(World Bank, 2006). As ICT adoption in industries becomes widespread, the demand
for workers with computer skills also has increased, generating higher wage rates for
ICT-savvy individuals (van Ark, O’Mahony and Timmer, 2008). As a result, skills
associated with ICT are crucial for individuals to be competitive in labor markets and
secure higher earnings (Autor, Katz and Krueger, 1998; Vicente and Lopez, 2011).
Unfortunately, the inequality of access to and expertise in ICT (henceforth, the
digital divide) is substantial in both developed and developing countries (Norris,
2001). In the US, 80 percent of the households with an income over $75,000 have
internet at home, while only 25 percent of the poorest households have the access
(Dickard and Schneider, 2011). In developing countries, such as India, the rate of
access to the internet of the urban households in 2008 was 10 times that of rural ones
(Singh, 2010). In China, internet penetration was almost 4 times higher in urban areas
than in rural areas (Chinese Internet Network Information Center, 2010). Computer
ownership is found to be 14 times higher for urban children than the rural children
(Yang et al., 2012). As access to ICT during childhood is a strong predictor of
expertise in ICT in adulthood, the school-aged digital divide may be expected to be
transformed into future disparities in productivity and earnings (Baouendi and Wilson,
1989). There is a concern that a persistent digital divide in society may contribute to
entrenched and long term stratification of wealth, opportunity and quality of life
between peoples of the world (Autor, Katz and Krueger, 1998).
The One Laptop per Child program (henceforth, OLPC) is one high profile
initiative that has aimed to narrow the digital divide. The OLPC concept was first
proposed by MIT professor and investor Nicholas Negroponte, who had a vision to
revolutionize education through the development and distribution of low-cost laptops
(Buchele and Owusu-Aning, 2007). OLPC was designed to bridge the digital divide
by providing inexpensive computers with network capability to poor children around
the world who would otherwise not likely have access to them. The laptops were
specially designed to cope with low-power supply and the ruggedness of poor urban
and rural areas. The software included with the machines consisted of a graphical user
interface and programs designed to improve learning by allowing interaction between
users and access to information via networking and the internet. The OLPC program
states that it was set up to provide a more efficient infrastructure for learning and
gaining access to information (Mangiatordi and Pischetola, 2010). The program was
also expected to generate enthusiasm for learning among students, improve
educational performance and help users overcome the digital divide (Bhatta, 2008).
A number of scholars, however, have disputed the premise that providing
laptops to disadvantaged children will either reduce the digital divide or improve
educational outcomes. For example, it is pointed out that without certain classroom
structure or teacher training, the program is almost certainly going to be ineffective in
achieving its goals (Butler, 2007). Furthermore, developing countries typically do not
have access to learning-based software and other digital content that sustain the
long-term interest of children, which is essential to OLPC’s success (Kraemer,
Dedrick and Sharma, 2009).
Given the controversy about OLPC’s efficacy and the fact that more than two
million OLPC laptops have been distributed in more than 40 countries (Verma, 2011),
there is surprisingly little empirical evidence on the impacts of the program. Most of
OLPC projects do not set up formal evaluations (Nugroho and Lonsdale, 2009). Even
if they do, the evaluations are either anecdotal or descriptive (Hourcade, 2008). The
OLPC project in Sri Lanka is one of the two projects that did incorporate systematic
data collection into its design. However, no official report is available so far
(Aturupane, 2010). To date, we have only found a descriptive study which states that
the OLPC program had a positive impact on math and English test scores based on
schooling grading registries in three of the sample schools of the OLPC project
(Mozelius, Rahuman and Wikramanayake, 2011). A study to evaluate the impact of
OLPC in Haiti was severely interrupted by the major earthquake there in 2008
Table 1. Comparisons of the student characteristics between the attrited students and those
remaining in the sample and the characteristics of attrited students between the treatment and
control group in the 13 migrant schools.
Sample:sample+attrition
obs. d
Sample:attrition
obs. e
Dependent variable: attrition (1=attrited;
0=remained)
Dependent variable: treatment
(1=treatment; 0=control)
[1] [2]
[1] Baseline math score (units of standard deviation) a 0.03
-0.08
(0.03)
(0.22) [2] Baseline Chinese score (units of standard deviation)
b -0.05*
0.24
(0.03)
(0.16) [3] Baseline computer skills (units of standard deviation)
c 0.03
0.22
(0.03)
(0.19) [4] Age (number of years) -0.11***
0.11
(0.03)
(0.17) [5] Male (1=yes;0=no) -0.05
0.43*
(0.04)
(0.21) [6] Baseline math study efficacy scale (1-4 points) -0.04
-0.13
(0.05)
(0.30) [7] Student used computer before (1=yes;0=no) 0.12**
-0.16
(0.05)
(0.30) [8] Student had access to internet (1=yes;0=no) 0.01
0.03
(0.06)
(0.31) [9] Student is an only child (1=yes;0=no) -0.09*
0.22
(0.05)
(0.69) [10] Age of father (number of years) 0.01
0.04
(0.01)
(0.03) [11] Age of mother (number of years) -0.01
-0.03
(0.01)
(0.03) [12] Father has a junior high school or higher degree (1=yes;0=no) -0.02
-0.02
(0.06)
(0.32) [13] Mother has a junior high school or higher degree (1=yes;0=no) -0.10*
-0.02
(0.05)
(0.24) [14] Father runs a business (1=yes;0=no) 0.14*
0.14
(0.08)
(0.33) [15] Mother runs a business (1=yes;0=no) -0.08
0.33
(0.09)
(0.36) [16] Class dummy variables Yes
Yes
[17] Observations 300
50 [18] R-squared 0.223 0.608
Source: Authors’ survey.
* significant at 10%, ** significant at 5%, ***significant at 1%. Robust standard errors in brackets. ab
The baseline math/Chinese score is the score on the standardized math/Chinese test that was given to
all students in the sample before the OLPC program. c The baseline computer skills scale is the standardized mean score on a set of computer skill questions
that was given to all students in the sample before the OLPC program. d The sample includes both the sample observations (non-attrition) and the attrition observations.
e The sample is limited to the attrited observations.
Table 2. Comparison of student and family characteristic between the treatment and control
groups in the 13 migrant schools.
Treatment
Control
Difference=
(128 obs.) (122 obs.) Treatment-Control
Mean SD
Mean SD
Mean P-value
[1] [2] [3] [4] [5] [6]
[1] Baseline math score (units of standard
deviation) a
0.16 0.97
-0.01 0.96
0.17 0.17
[2] Baseline Chinese score (units of
standard deviation) b
0.12 0.87
0.01 0.96
0.11 0.34
[3] Baseline computer skills score (units
of standard deviation) c
0.04 0.99
-0.06 1.05
0.10 0.44
[4] Age (number of years) 10.01 0.83
10.10 0.99
-0.09 0.43
[5] Male (1=yes;0=no) 0.56 0.50
0.57 0.50
-0.01 0.86
[6] Student had transferred to a different
school (1=yes;0=no) 0.14 0.35
0.20 0.40
0.06 0.24
[7] Baseline math study efficacy scale
(1-4 points) 3.34 0.46
3.31 0.51
0.03 0.58
[8] Student used computer before
(1=yes;0=no) 0.70 0.46
0.65 0.48
0.05 0.42
[9] Student had access to internet
(1=yes;0=no) 0.34 0.48
0.37 0.48
-0.02 0.68
[10] Student is an only child (1=yes;0=no) 0.19 0.39
0.22 0.42
-0.03 0.51
[11] Age of father (number of years) 37.41 4.97
36.94 4.82
0.47 0.45
[12] Age of mother (number of years) 35.52 3.96
35.59 4.65
-0.07 0.90
[13] Father has a junior high school or
higher degree (1=yes;0=no) 0.79 0.41
0.78 0.42
0.01 0.84
[14] Mother has a junior high school or
higher degree (1=yes;0=no) 0.65 0.48
0.66 0.48
-0.01 0.90
[15] Father runs a business (1=yes;0=no) 0.23 0.43
0.27 0.45
-0.04 0.51
[16] Mother runs a business (1=yes;0=no) 0.18 0.39 0.21 0.41 -0.03 0.51
Source: Authors’ survey. ab
The baseline math/Chinese score is the score on the standardized math/Chinese test that was given to
all students in the sample before the OLPC program. c The baseline computer skills scale is the standardized mean score on a set of computer skill questions
that was given to all students in the sample before the OLPC program.
Table 3. Ordinary Least Squares estimators of the impacts of OLPC program on student
[1] Treatment (1=treatment;0=control) 0.32** 0.33*** (0.12) (0.10) [2] Baseline math score (units of standard deviation)
a 0.09
(0.08) [3] Baseline Chinese score (units of standard deviation)
b
-0.07
(0.07) [4] Baseline computer skills score (units of standard
deviation) c
-0.67***
(0.07) [5] Age (number of years) -0.06 (0.06) [6] Male (1=yes;0=no) 0.23** (0.11) [7] Student had transferred to a different school
(1=yes;0=no)
-0.19 (0.16) [8] Baseline math study efficacy scale (1-4 points) 0.05 (0.11) [9] Student used computer before (1=yes;0=no) 0.28*** (0.11) [10] Student had access to internet (1=yes;0=no) -0.09 (0.14) [11] Student is an only child (1=yes;0=no) -0.01 (0.02) [12] Age of father (number of years) 0.01 (0.01) [13] Age of mother (number of years) -0.05 (0.13) [14] Father has a junior high school or higher degree
(1=yes;0=no)
0.01 (0.12) [15] Mother has a junior high school or higher degree
(1=yes;0=no)
0.27* (0.14) [16] Father runs a business (1=yes;0=no) -0.17 (0.17) [17] Mother runs a business (1=yes;0=no) 0.33*** (0.10) [18] Class dummy variables No Yes [19] Observations 250 250 [20] R-squared 0.026 0.545
Source: Authors’ survey.
* significant at 10%, ** significant at 5%, ***significant at 1%. Robust standard errors in brackets. ab
The baseline math/Chinese score is the score on the standardized math/Chinese test that was given to
all students in the sample before the OLPC program. c The baseline computer skills scale is the standardized mean score on a set of computer skill questions
that was given to all students in the sample before the OLPC program.
Table 4. Ordinary Least Squares estimators of the impacts of OLPC program on student
standardized math test scores in the 13 migrant schools.
Dependent variable: standardized post-OLPC math test score-standardized baseline math test score
(1) (2)
[1] Treatment (1=treatment;0=control) 0.07 0.17* (0.11) (0.10) [2] Baseline math score (units of standard deviation)
a -0.39***
(0.07) [3] Baseline Chinese score (units of standard deviation)
b 0.11
(0.07) [4] Baseline computer skills score (units of standard
deviation) c
-0.02
(0.07) [5] Age (number of years) -0.01 (0.07) [6] Male (1=yes;0=no) 0.21** (0.10) [7] Student had transferred to a different school
(1=yes;0=no) 0.23
(0.22) [8] Baseline math study efficacy scale (1-4 points) 0.09 (0.12) [9] Student used computer before (1=yes;0=no) 0.05 (0.14) [10] Student had access to internet (1=yes;0=no) 0.08 (0.12) [11] Student is an only child (1=yes;0=no) 0.08 (0.13) [12] Age of father (number of years) -0.00 (0.02) [13] Age of mother (number of years) 0.01 (0.02) [14] Father has a junior high school or higher degree
(1=yes;0=no) 0.06
(0.13) [15] Mother has a junior high school or higher degree
(1=yes;0=no) -0.10
(0.12) [16] Father runs a business (1=yes;0=no) 0.11 (0.18) [17] Mother runs a business (1=yes;0=no) -0.17 (0.21) [18] Class dummy variables No Yes [19] Observations 250 250 [20] R-squared 0.002 0.358
Source: Authors’ survey.
* significant at 10%, ** significant at 5%, ***significant at 1%. Robust standard errors in brackets. ab
The baseline math/Chinese score is the score on the standardized math/Chinese test that was given to
all students in the sample before the OLPC program. c The baseline computer skills scale is the standardized mean score on a set of computer skill questions
that was given to all students in the sample before the OLPC program.
Table 5. Ordinary Least Squares estimators of the impacts of OLPC program on student
standardized Chinese test scores in the 13 migrant schools.
Dependent variable: standardized post-OLPC Chinese test score-standardized baseline Chinese test score
(1) (2)
[1] Treatment (1=treatment;0=control) -0.03 0.01
(0.13) (0.12)
[2] Baseline math score (units of standard deviation) a
0.31***
(0.08)
[3] Baseline Chinese score (units of standard deviation) b
-0.65***
(0.08)
[4] Baseline computer skills score (units of standard deviation) c
-0.09
(0.07)
[5] Age (number of years)
0.05
(0.07)
[6] Male (1=yes;0=no)
-0.14
(0.12)
[7] Student had transferred to a different school (1=yes;0=no)
-0.10
(0.21)
[8] Baseline math study efficacy scale (1-4 points)
0.19
(0.13)
[9] Student used computer before (1=yes;0=no)
-0.03
(0.15)
[10]
0.02
(0.14)
[11] Student is an only child (1=yes;0=no)
0.29**
(0.12)
[12] Age of father (number of years)
0.00
(0.02)
[13] Age of mother (number of years)
0.01
(0.02)
[14] Father has a junior high school or higher degree (1=yes;0=no)
-0.21
(0.15)
[15] Mother has a junior high school or higher degree (1=yes;0=no)
0.36***
(0.13)
[16] Father runs a business (1=yes;0=no)
0.30
(0.20)
[17] Mother runs a business (1=yes;0=no)
-0.28
(0.22)
[18] Class dummy variables No Yes [19] Observations 250 250 [20] R-squared 0.000 0.447
Source: Authors’ survey.
* significant at 10%, ** significant at 5%, ***significant at 1%. Robust standard errors in brackets. ab
The baseline math/Chinese score is the score on the standardized math/Chinese test that was given to
all students in the sample before the OLPC program. c The baseline computer skills scale is the standardized mean score on a set of computer skill questions
that was given to all students in the sample before the OLPC program.
Table 6. The Ordinary Least Squares estimators of the heterogeneous program effect on
student computer skills, standardized math scores and standardized Chinese scores with
different characteristics in the 13 migrant schools.
* significant at 10%, ** significant at 5%, ***significant at 1%. Robust standard errors in brackets. ab
The baseline math/Chinese score is the score on the standardized math/Chinese test that was given to
all students in the sample before the OLPC program. c The baseline computer skills scale is the standardized mean score on a set of computer skill questions
that was given to all students in the sample before the OLPC program. d Control variables include all the variables that are included in Table 1.
Table 7. The Ordinary Least Squares estimators of the program effect on additional outcome
variables in the 13 migrant schools.
Dependent variables
Learning activity
using computers
(1=used any
computer software
for learning; 0=did
not use computer
software for
learning)
TV watching
(1=watched TV one
day before the survey;
0=did not watch TV
one day before the
survey)
Student
self-esteem scale
(1-4 points)
(1) (3) (4)
[1] Treatment
(1=treatment;0=control)
0.14** -0.12* 0.12**
(0.07) (0.06) (0.05)
[2] Control variables a Yes Yes Yes
[3] Class dummy variables Yes Yes Yes
[4] Observations 250 250 250
[5] R-squared 0.267 0.200 0.256
* significant at 10%, ** significant at 5%, ***significant at 1%. Robust standard errors in brackets. a Control variables include all the variables that are included in Table 1.
Figure 1. Experiment profile
Baseline (Dec. 2010)
794 grade 3 students in 13 migrant
schools in the greater Beijing area
Randomization (Apr. 2011)
Evaluation (Oct. 2011)
Randomly selected RCT
participants: 300 students
Non-RCT participants
Randomly assigned 150
students to the treatment group
Randomly assigned 150
students to the control group
Attrition: 22 students Attrition: 28 students
Sample analyzed: 128 in the treatment group; 122 in the
control group
Figure 2. Change in the standardized computer skill scale before and after the OLPC program
Panel A. Standardized computer skill scale before and after OLPC: the treatment and control
group.
Panel B. Difference in difference in the standardized computer skill scale before and after the
OLPC Program between the treatment and control group.
Figure 3. Change in the standardized math test scores before and after the OLPC program
Panel A. Standardized math test scores before and after OLPC: the treatment and control
group.
Panel B. Difference in difference in the standardized math test scores before and after the
OLPC Program between the treatment and control group.
Figure 4. Change in the standardized Chinese test scores before and after the OLPC program
Panel A. Standardized Chinese test scores before and after OLPC: the treatment and control
group.
Panel B. Difference in difference in the standardized Chinese test scores before and after the
OLPC Program between the treatment and control group.